2016-02-10 00:12:00 +01:00
|
|
|
// Copyright (c) 2011-present, Facebook, Inc. All rights reserved.
|
2017-07-16 01:03:42 +02:00
|
|
|
// This source code is licensed under both the GPLv2 (found in the
|
|
|
|
// COPYING file in the root directory) and Apache 2.0 License
|
|
|
|
// (found in the LICENSE.Apache file in the root directory).
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
//
|
|
|
|
// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
|
|
|
|
// Use of this source code is governed by a BSD-style license that can be
|
|
|
|
// found in the LICENSE file. See the AUTHORS file for names of contributors.
|
|
|
|
|
2014-04-15 22:39:26 +02:00
|
|
|
#ifndef ROCKSDB_LITE
|
|
|
|
|
2020-06-25 04:30:15 +02:00
|
|
|
#include "rocksdb/utilities/backupable_db.h"
|
|
|
|
|
2015-01-04 05:02:52 +01:00
|
|
|
#include <stdlib.h>
|
2020-06-25 04:30:15 +02:00
|
|
|
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
#include <algorithm>
|
2016-06-10 04:03:10 +02:00
|
|
|
#include <atomic>
|
2019-09-20 21:00:55 +02:00
|
|
|
#include <cinttypes>
|
2016-12-16 20:17:26 +01:00
|
|
|
#include <functional>
|
2016-06-10 04:03:10 +02:00
|
|
|
#include <future>
|
|
|
|
#include <limits>
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
#include <map>
|
2015-08-25 22:32:46 +02:00
|
|
|
#include <mutex>
|
2015-02-27 23:03:56 +01:00
|
|
|
#include <sstream>
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
#include <string>
|
2015-07-02 20:35:51 +02:00
|
|
|
#include <thread>
|
2014-01-25 01:15:05 +01:00
|
|
|
#include <unordered_map>
|
2015-07-02 20:35:51 +02:00
|
|
|
#include <unordered_set>
|
2016-06-10 04:03:10 +02:00
|
|
|
#include <vector>
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
|
2020-08-12 22:28:43 +02:00
|
|
|
#include "db/log_reader.h"
|
Introduce a new storage specific Env API (#5761)
Summary:
The current Env API encompasses both storage/file operations, as well as OS related operations. Most of the APIs return a Status, which does not have enough metadata about an error, such as whether its retry-able or not, scope (i.e fault domain) of the error etc., that may be required in order to properly handle a storage error. The file APIs also do not provide enough control over the IO SLA, such as timeout, prioritization, hinting about placement and redundancy etc.
This PR separates out the file/storage APIs from Env into a new FileSystem class. The APIs are updated to return an IOStatus with metadata about the error, as well as to take an IOOptions structure as input in order to allow more control over the IO.
The user can set both ```options.env``` and ```options.file_system``` to specify that RocksDB should use the former for OS related operations and the latter for storage operations. Internally, a ```CompositeEnvWrapper``` has been introduced that inherits from ```Env``` and redirects individual methods to either an ```Env``` implementation or the ```FileSystem``` as appropriate. When options are sanitized during ```DB::Open```, ```options.env``` is replaced with a newly allocated ```CompositeEnvWrapper``` instance if both env and file_system have been specified. This way, the rest of the RocksDB code can continue to function as before.
This PR also ports PosixEnv to the new API by splitting it into two - PosixEnv and PosixFileSystem. PosixEnv is defined as a sub-class of CompositeEnvWrapper, and threading/time functions are overridden with Posix specific implementations in order to avoid an extra level of indirection.
The ```CompositeEnvWrapper``` translates ```IOStatus``` return code to ```Status```, and sets the severity to ```kSoftError``` if the io_status is retryable. The error handling code in RocksDB can then recover the DB automatically.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5761
Differential Revision: D18868376
Pulled By: anand1976
fbshipit-source-id: 39efe18a162ea746fabac6360ff529baba48486f
2019-12-13 23:47:08 +01:00
|
|
|
#include "env/composite_env_wrapper.h"
|
Auto-GarbageCollect on PurgeOldBackups and DeleteBackup (#6015)
Summary:
Only if there is a crash, power failure, or I/O error in
DeleteBackup, shared or private files from the backup might be left
behind that are not cleaned up by PurgeOldBackups or DeleteBackup-- only
by GarbageCollect. This makes the BackupEngine API "leaky by default."
Even if it means a modest performance hit, I think we should make
Delete and Purge do as they say, with ongoing best effort: i.e. future
calls will attempt to finish any incomplete work from earlier calls.
This change does that by having DeleteBackup and PurgeOldBackups do a
GarbageCollect, unless (to minimize performance hit) this BackupEngine
has already done a GarbageCollect and there have been no
deletion-related I/O errors in that GarbageCollect or since then.
Rejected alternative 1: remove meta file last instead of first. This would in theory turn partially deleted backups into corrupted backups, but code changes would be needed to allow the missing files and consider it acceptably corrupt, rather than failing to open the BackupEngine. This might be a reasonable choice, but I mostly rejected it because it doesn't solve the legacy problem of cleaning up existing lingering files.
Rejected alternative 2: use a deletion marker file. If deletion started with creating a file that marks a backup as flagged for deletion, then we could reliably detect partially deleted backups and efficiently finish removing them. In addition to not solving the legacy problem, this could be precarious if there's a disk full situation, and we try to create a new file in order to delete some files. Ugh.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6015
Test Plan: Updated unit tests
Differential Revision: D18401333
Pulled By: pdillinger
fbshipit-source-id: 12944e372ce6809f3f5a4c416c3b321a8927d925
2019-11-09 04:13:41 +01:00
|
|
|
#include "file/filename.h"
|
|
|
|
#include "file/sequence_file_reader.h"
|
|
|
|
#include "file/writable_file_writer.h"
|
|
|
|
#include "logging/logging.h"
|
|
|
|
#include "port/port.h"
|
|
|
|
#include "rocksdb/rate_limiter.h"
|
|
|
|
#include "rocksdb/transaction_log.h"
|
2020-06-25 04:30:15 +02:00
|
|
|
#include "table/sst_file_dumper.h"
|
Auto-GarbageCollect on PurgeOldBackups and DeleteBackup (#6015)
Summary:
Only if there is a crash, power failure, or I/O error in
DeleteBackup, shared or private files from the backup might be left
behind that are not cleaned up by PurgeOldBackups or DeleteBackup-- only
by GarbageCollect. This makes the BackupEngine API "leaky by default."
Even if it means a modest performance hit, I think we should make
Delete and Purge do as they say, with ongoing best effort: i.e. future
calls will attempt to finish any incomplete work from earlier calls.
This change does that by having DeleteBackup and PurgeOldBackups do a
GarbageCollect, unless (to minimize performance hit) this BackupEngine
has already done a GarbageCollect and there have been no
deletion-related I/O errors in that GarbageCollect or since then.
Rejected alternative 1: remove meta file last instead of first. This would in theory turn partially deleted backups into corrupted backups, but code changes would be needed to allow the missing files and consider it acceptably corrupt, rather than failing to open the BackupEngine. This might be a reasonable choice, but I mostly rejected it because it doesn't solve the legacy problem of cleaning up existing lingering files.
Rejected alternative 2: use a deletion marker file. If deletion started with creating a file that marks a backup as flagged for deletion, then we could reliably detect partially deleted backups and efficiently finish removing them. In addition to not solving the legacy problem, this could be precarious if there's a disk full situation, and we try to create a new file in order to delete some files. Ugh.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6015
Test Plan: Updated unit tests
Differential Revision: D18401333
Pulled By: pdillinger
fbshipit-source-id: 12944e372ce6809f3f5a4c416c3b321a8927d925
2019-11-09 04:13:41 +01:00
|
|
|
#include "test_util/sync_point.h"
|
|
|
|
#include "util/channel.h"
|
|
|
|
#include "util/coding.h"
|
|
|
|
#include "util/crc32c.h"
|
|
|
|
#include "util/string_util.h"
|
|
|
|
#include "utilities/checkpoint/checkpoint_impl.h"
|
|
|
|
|
2020-02-20 21:07:53 +01:00
|
|
|
namespace ROCKSDB_NAMESPACE {
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
namespace {
|
|
|
|
inline uint32_t ChecksumHexToInt32(const std::string& checksum_hex) {
|
|
|
|
std::string checksum_str;
|
|
|
|
Slice(checksum_hex).DecodeHex(&checksum_str);
|
|
|
|
return EndianSwapValue(DecodeFixed32(checksum_str.c_str()));
|
|
|
|
}
|
|
|
|
inline std::string ChecksumStrToHex(const std::string& checksum_str) {
|
|
|
|
return Slice(checksum_str).ToString(true);
|
|
|
|
}
|
|
|
|
inline std::string ChecksumInt32ToHex(const uint32_t& checksum_value) {
|
|
|
|
std::string checksum_str;
|
|
|
|
PutFixed32(&checksum_str, EndianSwapValue(checksum_value));
|
|
|
|
return ChecksumStrToHex(checksum_str);
|
|
|
|
}
|
2020-08-12 22:28:43 +02:00
|
|
|
// Checks if the checksum function names are the same. Note that both the
|
|
|
|
// backup default checksum function and the db standard checksum function are
|
|
|
|
// crc32c although they have different names. So We treat the db standard
|
|
|
|
// checksum function name and the backup default checksum function name as
|
|
|
|
// the same name.
|
|
|
|
inline bool IsSameChecksumFunc(const std::string& dst_checksum_func_name,
|
|
|
|
const std::string& src_checksum_func_name) {
|
|
|
|
return (dst_checksum_func_name == src_checksum_func_name) ||
|
|
|
|
((dst_checksum_func_name == kDefaultBackupFileChecksumFuncName) &&
|
|
|
|
(src_checksum_func_name == kStandardDbFileChecksumFuncName)) ||
|
|
|
|
((src_checksum_func_name == kDefaultBackupFileChecksumFuncName) &&
|
|
|
|
(dst_checksum_func_name == kStandardDbFileChecksumFuncName));
|
|
|
|
}
|
|
|
|
inline bool IsSstFile(const std::string& fname) {
|
|
|
|
return fname.length() > 4 && fname.rfind(".sst") == fname.length() - 4;
|
|
|
|
}
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
} // namespace
|
|
|
|
|
2014-09-09 22:44:42 +02:00
|
|
|
void BackupStatistics::IncrementNumberSuccessBackup() {
|
|
|
|
number_success_backup++;
|
|
|
|
}
|
|
|
|
void BackupStatistics::IncrementNumberFailBackup() {
|
|
|
|
number_fail_backup++;
|
|
|
|
}
|
|
|
|
|
|
|
|
uint32_t BackupStatistics::GetNumberSuccessBackup() const {
|
|
|
|
return number_success_backup;
|
|
|
|
}
|
|
|
|
uint32_t BackupStatistics::GetNumberFailBackup() const {
|
|
|
|
return number_fail_backup;
|
|
|
|
}
|
|
|
|
|
|
|
|
std::string BackupStatistics::ToString() const {
|
|
|
|
char result[50];
|
|
|
|
snprintf(result, sizeof(result), "# success backup: %u, # fail backup: %u",
|
|
|
|
GetNumberSuccessBackup(), GetNumberFailBackup());
|
|
|
|
return result;
|
|
|
|
}
|
|
|
|
|
2014-03-10 20:09:54 +01:00
|
|
|
void BackupableDBOptions::Dump(Logger* logger) const {
|
2017-03-16 03:22:52 +01:00
|
|
|
ROCKS_LOG_INFO(logger, " Options.backup_dir: %s",
|
|
|
|
backup_dir.c_str());
|
|
|
|
ROCKS_LOG_INFO(logger, " Options.backup_env: %p", backup_env);
|
|
|
|
ROCKS_LOG_INFO(logger, " Options.share_table_files: %d",
|
|
|
|
static_cast<int>(share_table_files));
|
|
|
|
ROCKS_LOG_INFO(logger, " Options.info_log: %p", info_log);
|
|
|
|
ROCKS_LOG_INFO(logger, " Options.sync: %d",
|
|
|
|
static_cast<int>(sync));
|
|
|
|
ROCKS_LOG_INFO(logger, " Options.destroy_old_data: %d",
|
|
|
|
static_cast<int>(destroy_old_data));
|
|
|
|
ROCKS_LOG_INFO(logger, " Options.backup_log_files: %d",
|
|
|
|
static_cast<int>(backup_log_files));
|
|
|
|
ROCKS_LOG_INFO(logger, " Options.backup_rate_limit: %" PRIu64,
|
|
|
|
backup_rate_limit);
|
|
|
|
ROCKS_LOG_INFO(logger, " Options.restore_rate_limit: %" PRIu64,
|
|
|
|
restore_rate_limit);
|
|
|
|
ROCKS_LOG_INFO(logger, "Options.max_background_operations: %d",
|
|
|
|
max_background_operations);
|
2014-03-10 20:09:54 +01:00
|
|
|
}
|
|
|
|
|
2014-01-28 20:26:07 +01:00
|
|
|
// -------- BackupEngineImpl class ---------
|
|
|
|
class BackupEngineImpl : public BackupEngine {
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
public:
|
2020-03-29 04:05:54 +02:00
|
|
|
BackupEngineImpl(const BackupableDBOptions& options, Env* db_env,
|
2014-04-25 21:49:29 +02:00
|
|
|
bool read_only = false);
|
2019-02-14 22:52:47 +01:00
|
|
|
~BackupEngineImpl() override;
|
2020-03-29 04:05:54 +02:00
|
|
|
|
|
|
|
using BackupEngine::CreateNewBackupWithMetadata;
|
|
|
|
Status CreateNewBackupWithMetadata(const CreateBackupOptions& options, DB* db,
|
|
|
|
const std::string& app_metadata) override;
|
|
|
|
|
2015-02-26 20:28:41 +01:00
|
|
|
Status PurgeOldBackups(uint32_t num_backups_to_keep) override;
|
2020-03-29 04:05:54 +02:00
|
|
|
|
2015-02-26 20:28:41 +01:00
|
|
|
Status DeleteBackup(BackupID backup_id) override;
|
2020-03-29 04:05:54 +02:00
|
|
|
|
2015-02-26 20:28:41 +01:00
|
|
|
void StopBackup() override {
|
2014-01-09 21:24:28 +01:00
|
|
|
stop_backup_.store(true, std::memory_order_release);
|
|
|
|
}
|
2020-03-29 04:05:54 +02:00
|
|
|
|
2015-02-26 20:28:41 +01:00
|
|
|
Status GarbageCollect() override;
|
|
|
|
|
2016-07-06 21:54:01 +02:00
|
|
|
// The returned BackupInfos are in chronological order, which means the
|
|
|
|
// latest backup comes last.
|
2015-02-26 20:28:41 +01:00
|
|
|
void GetBackupInfo(std::vector<BackupInfo>* backup_info) override;
|
2020-03-29 04:05:54 +02:00
|
|
|
|
2015-02-26 20:28:41 +01:00
|
|
|
void GetCorruptedBackups(std::vector<BackupID>* corrupt_backup_ids) override;
|
2020-03-29 04:05:54 +02:00
|
|
|
|
|
|
|
using BackupEngine::RestoreDBFromBackup;
|
|
|
|
Status RestoreDBFromBackup(const RestoreOptions& options, BackupID backup_id,
|
|
|
|
const std::string& db_dir,
|
|
|
|
const std::string& wal_dir) override;
|
|
|
|
|
|
|
|
using BackupEngine::RestoreDBFromLatestBackup;
|
|
|
|
Status RestoreDBFromLatestBackup(const RestoreOptions& options,
|
|
|
|
const std::string& db_dir,
|
|
|
|
const std::string& wal_dir) override {
|
|
|
|
return RestoreDBFromBackup(options, latest_valid_backup_id_, db_dir,
|
|
|
|
wal_dir);
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
}
|
|
|
|
|
2020-06-26 20:39:43 +02:00
|
|
|
Status VerifyBackup(BackupID backup_id,
|
2020-07-01 03:45:36 +02:00
|
|
|
bool verify_with_checksum = false) override;
|
2015-09-02 18:03:55 +02:00
|
|
|
|
2015-07-14 20:51:36 +02:00
|
|
|
Status Initialize();
|
|
|
|
|
2020-07-01 03:45:36 +02:00
|
|
|
// Obtain the naming option for backup table files
|
|
|
|
BackupTableNameOption GetTableNamingOption() const {
|
|
|
|
return options_.share_files_with_checksum_naming;
|
|
|
|
}
|
2020-06-25 04:30:15 +02:00
|
|
|
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
private:
|
2014-03-17 23:39:23 +01:00
|
|
|
void DeleteChildren(const std::string& dir, uint32_t file_type_filter = 0);
|
Auto-GarbageCollect on PurgeOldBackups and DeleteBackup (#6015)
Summary:
Only if there is a crash, power failure, or I/O error in
DeleteBackup, shared or private files from the backup might be left
behind that are not cleaned up by PurgeOldBackups or DeleteBackup-- only
by GarbageCollect. This makes the BackupEngine API "leaky by default."
Even if it means a modest performance hit, I think we should make
Delete and Purge do as they say, with ongoing best effort: i.e. future
calls will attempt to finish any incomplete work from earlier calls.
This change does that by having DeleteBackup and PurgeOldBackups do a
GarbageCollect, unless (to minimize performance hit) this BackupEngine
has already done a GarbageCollect and there have been no
deletion-related I/O errors in that GarbageCollect or since then.
Rejected alternative 1: remove meta file last instead of first. This would in theory turn partially deleted backups into corrupted backups, but code changes would be needed to allow the missing files and consider it acceptably corrupt, rather than failing to open the BackupEngine. This might be a reasonable choice, but I mostly rejected it because it doesn't solve the legacy problem of cleaning up existing lingering files.
Rejected alternative 2: use a deletion marker file. If deletion started with creating a file that marks a backup as flagged for deletion, then we could reliably detect partially deleted backups and efficiently finish removing them. In addition to not solving the legacy problem, this could be precarious if there's a disk full situation, and we try to create a new file in order to delete some files. Ugh.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6015
Test Plan: Updated unit tests
Differential Revision: D18401333
Pulled By: pdillinger
fbshipit-source-id: 12944e372ce6809f3f5a4c416c3b321a8927d925
2019-11-09 04:13:41 +01:00
|
|
|
Status DeleteBackupInternal(BackupID backup_id);
|
2014-03-17 23:39:23 +01:00
|
|
|
|
2016-03-02 04:33:33 +01:00
|
|
|
// Extends the "result" map with pathname->size mappings for the contents of
|
2016-04-26 21:33:30 +02:00
|
|
|
// "dir" in "env". Pathnames are prefixed with "dir".
|
2016-03-02 04:33:33 +01:00
|
|
|
Status InsertPathnameToSizeBytes(
|
2016-04-26 21:33:30 +02:00
|
|
|
const std::string& dir, Env* env,
|
2016-03-02 04:33:33 +01:00
|
|
|
std::unordered_map<std::string, uint64_t>* result);
|
|
|
|
|
2014-01-28 18:43:36 +01:00
|
|
|
struct FileInfo {
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
FileInfo(const std::string& fname, uint64_t sz, const std::string& checksum,
|
2020-08-12 22:28:43 +02:00
|
|
|
const std::string& custom_checksum,
|
|
|
|
const std::string& checksum_name, const std::string& id = "",
|
|
|
|
const std::string& sid = "")
|
2020-06-25 04:30:15 +02:00
|
|
|
: refs(0),
|
|
|
|
filename(fname),
|
|
|
|
size(sz),
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
checksum_hex(checksum),
|
2020-08-12 22:28:43 +02:00
|
|
|
custom_checksum_hex(custom_checksum),
|
|
|
|
checksum_func_name(checksum_name),
|
2020-06-25 04:30:15 +02:00
|
|
|
db_id(id),
|
|
|
|
db_session_id(sid) {}
|
2014-01-28 18:43:36 +01:00
|
|
|
|
2015-01-05 14:35:09 +01:00
|
|
|
FileInfo(const FileInfo&) = delete;
|
|
|
|
FileInfo& operator=(const FileInfo&) = delete;
|
|
|
|
|
2014-01-28 18:43:36 +01:00
|
|
|
int refs;
|
|
|
|
const std::string filename;
|
|
|
|
const uint64_t size;
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
const std::string checksum_hex;
|
2020-08-12 22:28:43 +02:00
|
|
|
const std::string custom_checksum_hex;
|
|
|
|
const std::string checksum_func_name;
|
2020-06-25 04:30:15 +02:00
|
|
|
// DB identities
|
|
|
|
// db_id is obtained for potential usage in the future but not used
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
// currently
|
2020-06-25 04:30:15 +02:00
|
|
|
const std::string db_id;
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
// db_session_id appears in the backup SST filename if the table naming
|
|
|
|
// option is kOptionalChecksumAndDbSessionId
|
2020-06-25 04:30:15 +02:00
|
|
|
const std::string db_session_id;
|
2014-01-28 18:43:36 +01:00
|
|
|
};
|
|
|
|
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
class BackupMeta {
|
|
|
|
public:
|
2018-02-22 02:33:14 +01:00
|
|
|
BackupMeta(
|
|
|
|
const std::string& meta_filename, const std::string& meta_tmp_filename,
|
2015-01-05 14:35:09 +01:00
|
|
|
std::unordered_map<std::string, std::shared_ptr<FileInfo>>* file_infos,
|
|
|
|
Env* env)
|
2018-02-22 02:33:14 +01:00
|
|
|
: timestamp_(0),
|
|
|
|
sequence_number_(0),
|
|
|
|
size_(0),
|
|
|
|
meta_filename_(meta_filename),
|
|
|
|
meta_tmp_filename_(meta_tmp_filename),
|
|
|
|
file_infos_(file_infos),
|
|
|
|
env_(env) {}
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
|
2015-01-05 14:35:09 +01:00
|
|
|
BackupMeta(const BackupMeta&) = delete;
|
|
|
|
BackupMeta& operator=(const BackupMeta&) = delete;
|
|
|
|
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
~BackupMeta() {}
|
|
|
|
|
|
|
|
void RecordTimestamp() {
|
|
|
|
env_->GetCurrentTime(×tamp_);
|
|
|
|
}
|
|
|
|
int64_t GetTimestamp() const {
|
|
|
|
return timestamp_;
|
|
|
|
}
|
|
|
|
uint64_t GetSize() const {
|
|
|
|
return size_;
|
|
|
|
}
|
2014-11-11 22:47:22 +01:00
|
|
|
uint32_t GetNumberFiles() { return static_cast<uint32_t>(files_.size()); }
|
2013-12-11 05:49:28 +01:00
|
|
|
void SetSequenceNumber(uint64_t sequence_number) {
|
|
|
|
sequence_number_ = sequence_number;
|
|
|
|
}
|
|
|
|
uint64_t GetSequenceNumber() {
|
|
|
|
return sequence_number_;
|
|
|
|
}
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
|
2016-04-01 19:56:52 +02:00
|
|
|
const std::string& GetAppMetadata() const { return app_metadata_; }
|
|
|
|
|
|
|
|
void SetAppMetadata(const std::string& app_metadata) {
|
|
|
|
app_metadata_ = app_metadata;
|
|
|
|
}
|
|
|
|
|
2015-01-05 14:35:09 +01:00
|
|
|
Status AddFile(std::shared_ptr<FileInfo> file_info);
|
2014-01-28 18:43:36 +01:00
|
|
|
|
2015-07-14 20:51:36 +02:00
|
|
|
Status Delete(bool delete_meta = true);
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
|
|
|
|
bool Empty() {
|
|
|
|
return files_.empty();
|
|
|
|
}
|
|
|
|
|
2015-01-05 14:35:09 +01:00
|
|
|
std::shared_ptr<FileInfo> GetFile(const std::string& filename) const {
|
2015-01-04 05:06:59 +01:00
|
|
|
auto it = file_infos_->find(filename);
|
|
|
|
if (it == file_infos_->end())
|
|
|
|
return nullptr;
|
2015-01-05 14:35:09 +01:00
|
|
|
return it->second;
|
2015-01-04 05:06:59 +01:00
|
|
|
}
|
|
|
|
|
2015-01-05 14:35:09 +01:00
|
|
|
const std::vector<std::shared_ptr<FileInfo>>& GetFiles() {
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
return files_;
|
|
|
|
}
|
|
|
|
|
2016-03-02 04:33:33 +01:00
|
|
|
// @param abs_path_to_size Pre-fetched file sizes (bytes).
|
|
|
|
Status LoadFromFile(
|
2016-03-03 22:32:20 +01:00
|
|
|
const std::string& backup_dir,
|
2016-03-02 04:33:33 +01:00
|
|
|
const std::unordered_map<std::string, uint64_t>& abs_path_to_size);
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
Status StoreToFile(bool sync);
|
|
|
|
|
2015-02-27 23:03:56 +01:00
|
|
|
std::string GetInfoString() {
|
|
|
|
std::ostringstream ss;
|
|
|
|
ss << "Timestamp: " << timestamp_ << std::endl;
|
|
|
|
char human_size[16];
|
|
|
|
AppendHumanBytes(size_, human_size, sizeof(human_size));
|
|
|
|
ss << "Size: " << human_size << std::endl;
|
|
|
|
ss << "Files:" << std::endl;
|
|
|
|
for (const auto& file : files_) {
|
|
|
|
AppendHumanBytes(file->size, human_size, sizeof(human_size));
|
|
|
|
ss << file->filename << ", size " << human_size << ", refs "
|
|
|
|
<< file->refs << std::endl;
|
|
|
|
}
|
|
|
|
return ss.str();
|
|
|
|
}
|
|
|
|
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
private:
|
|
|
|
int64_t timestamp_;
|
2013-12-11 05:49:28 +01:00
|
|
|
// sequence number is only approximate, should not be used
|
|
|
|
// by clients
|
|
|
|
uint64_t sequence_number_;
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
uint64_t size_;
|
2016-04-01 19:56:52 +02:00
|
|
|
std::string app_metadata_;
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
std::string const meta_filename_;
|
2018-02-22 02:33:14 +01:00
|
|
|
std::string const meta_tmp_filename_;
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
// files with relative paths (without "/" prefix!!)
|
2015-01-05 14:35:09 +01:00
|
|
|
std::vector<std::shared_ptr<FileInfo>> files_;
|
|
|
|
std::unordered_map<std::string, std::shared_ptr<FileInfo>>* file_infos_;
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
Env* env_;
|
2013-12-10 19:35:06 +01:00
|
|
|
|
2014-05-03 02:08:55 +02:00
|
|
|
static const size_t max_backup_meta_file_size_ = 10 * 1024 * 1024; // 10MB
|
|
|
|
}; // BackupMeta
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
|
|
|
|
inline std::string GetAbsolutePath(
|
|
|
|
const std::string &relative_path = "") const {
|
|
|
|
assert(relative_path.size() == 0 || relative_path[0] != '/');
|
|
|
|
return options_.backup_dir + "/" + relative_path;
|
|
|
|
}
|
|
|
|
inline std::string GetPrivateDirRel() const {
|
|
|
|
return "private";
|
|
|
|
}
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
inline std::string GetSharedDirRel() const { return "shared"; }
|
2014-05-03 02:08:55 +02:00
|
|
|
inline std::string GetSharedChecksumDirRel() const {
|
|
|
|
return "shared_checksum";
|
|
|
|
}
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
inline std::string GetPrivateFileRel(BackupID backup_id,
|
2014-01-09 21:24:28 +01:00
|
|
|
bool tmp = false,
|
|
|
|
const std::string& file = "") const {
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
assert(file.size() == 0 || file[0] != '/');
|
2020-02-20 21:07:53 +01:00
|
|
|
return GetPrivateDirRel() + "/" + ROCKSDB_NAMESPACE::ToString(backup_id) +
|
2014-01-09 21:24:28 +01:00
|
|
|
(tmp ? ".tmp" : "") + "/" + file;
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
}
|
2014-01-09 21:24:28 +01:00
|
|
|
inline std::string GetSharedFileRel(const std::string& file = "",
|
|
|
|
bool tmp = false) const {
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
assert(file.size() == 0 || file[0] != '/');
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
return GetSharedDirRel() + "/" + (tmp ? "." : "") + file +
|
2018-02-22 02:33:14 +01:00
|
|
|
(tmp ? ".tmp" : "");
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
}
|
2014-05-03 02:08:55 +02:00
|
|
|
inline std::string GetSharedFileWithChecksumRel(const std::string& file = "",
|
|
|
|
bool tmp = false) const {
|
|
|
|
assert(file.size() == 0 || file[0] != '/');
|
2018-02-22 02:33:14 +01:00
|
|
|
return GetSharedChecksumDirRel() + "/" + (tmp ? "." : "") + file +
|
|
|
|
(tmp ? ".tmp" : "");
|
2014-05-03 02:08:55 +02:00
|
|
|
}
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
inline bool UseSessionId(const std::string& sid) const {
|
|
|
|
return GetTableNamingOption() == kOptionalChecksumAndDbSessionId &&
|
|
|
|
!sid.empty();
|
|
|
|
}
|
2020-06-25 04:30:15 +02:00
|
|
|
inline std::string GetSharedFileWithChecksum(
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
const std::string& file, bool has_checksum,
|
|
|
|
const std::string& checksum_hex, const uint64_t file_size,
|
|
|
|
const std::string& db_session_id) const {
|
2014-05-03 02:08:55 +02:00
|
|
|
assert(file.size() == 0 || file[0] != '/');
|
|
|
|
std::string file_copy = file;
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
if (UseSessionId(db_session_id)) {
|
|
|
|
if (has_checksum) {
|
|
|
|
return file_copy.insert(file_copy.find_last_of('.'),
|
|
|
|
"_" + checksum_hex + "_" + db_session_id);
|
|
|
|
} else {
|
|
|
|
return file_copy.insert(file_copy.find_last_of('.'),
|
|
|
|
"_" + db_session_id);
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
return file_copy.insert(file_copy.find_last_of('.'),
|
|
|
|
"_" + ToString(ChecksumHexToInt32(checksum_hex)) +
|
|
|
|
"_" + ToString(file_size));
|
|
|
|
}
|
2014-05-03 02:08:55 +02:00
|
|
|
}
|
|
|
|
inline std::string GetFileFromChecksumFile(const std::string& file) const {
|
|
|
|
assert(file.size() == 0 || file[0] != '/');
|
|
|
|
std::string file_copy = file;
|
|
|
|
size_t first_underscore = file_copy.find_first_of('_');
|
|
|
|
return file_copy.erase(first_underscore,
|
|
|
|
file_copy.find_last_of('.') - first_underscore);
|
|
|
|
}
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
inline std::string GetBackupMetaDir() const {
|
|
|
|
return GetAbsolutePath("meta");
|
|
|
|
}
|
2018-02-22 02:33:14 +01:00
|
|
|
inline std::string GetBackupMetaFile(BackupID backup_id, bool tmp) const {
|
|
|
|
return GetBackupMetaDir() + "/" + (tmp ? "." : "") +
|
2020-02-20 21:07:53 +01:00
|
|
|
ROCKSDB_NAMESPACE::ToString(backup_id) + (tmp ? ".tmp" : "");
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
}
|
2020-08-12 22:28:43 +02:00
|
|
|
inline Status GetFileNameInfo(const std::string& file,
|
|
|
|
std::string& local_name, uint64_t& number,
|
|
|
|
FileType& type) const {
|
|
|
|
// 1. extract the filename
|
|
|
|
size_t last_slash = file.find_last_of('/');
|
|
|
|
// file will either be shared/<file>, shared_checksum/<file_crc32c_size>,
|
|
|
|
// shared_checksum/<file_session>, shared_checksum/<file_crc32c_session>,
|
|
|
|
// or private/<number>/<file>
|
|
|
|
assert(last_slash != std::string::npos);
|
|
|
|
local_name = file.substr(last_slash + 1);
|
|
|
|
|
|
|
|
// if the file was in shared_checksum, extract the real file name
|
|
|
|
// in this case the file is <number>_<checksum>_<size>.<type>,
|
|
|
|
// <number>_<session>.<type>, or <number>_<checksum>_<session>.<type>
|
|
|
|
if (file.substr(0, last_slash) == GetSharedChecksumDirRel()) {
|
|
|
|
local_name = GetFileFromChecksumFile(local_name);
|
|
|
|
}
|
|
|
|
|
|
|
|
// 2. find the filetype
|
|
|
|
bool ok = ParseFileName(local_name, &number, &type);
|
|
|
|
if (!ok) {
|
|
|
|
return Status::Corruption("Backup corrupted: Fail to parse filename " +
|
|
|
|
local_name);
|
|
|
|
}
|
|
|
|
return Status::OK();
|
|
|
|
}
|
|
|
|
inline bool HasCustomChecksumGenFactory() const {
|
|
|
|
return options_.file_checksum_gen_factory != nullptr;
|
|
|
|
}
|
|
|
|
// Returns nullptr if file_checksum_gen_factory is not set or
|
|
|
|
// file_checksum_gen_factory is not able to create a generator with
|
|
|
|
// name being requested_checksum_func_name
|
|
|
|
inline std::unique_ptr<FileChecksumGenerator> GetCustomChecksumGenerator(
|
|
|
|
const std::string& requested_checksum_func_name = "") const {
|
|
|
|
std::shared_ptr<FileChecksumGenFactory> checksum_factory =
|
|
|
|
options_.file_checksum_gen_factory;
|
|
|
|
if (checksum_factory == nullptr) {
|
|
|
|
return nullptr;
|
|
|
|
} else {
|
|
|
|
FileChecksumGenContext gen_context;
|
|
|
|
gen_context.requested_checksum_func_name = requested_checksum_func_name;
|
|
|
|
return checksum_factory->CreateFileChecksumGenerator(gen_context);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
// Set the checksum generator by the requested checksum function name
|
|
|
|
inline Status SetChecksumGenerator(
|
|
|
|
const std::string& requested_checksum_func_name,
|
|
|
|
std::unique_ptr<FileChecksumGenerator>& checksum_func) {
|
|
|
|
if (requested_checksum_func_name != kDefaultBackupFileChecksumFuncName) {
|
|
|
|
if (!HasCustomChecksumGenFactory()) {
|
|
|
|
// No custom checksum factory indicates users would like to use the
|
|
|
|
// backup default checksum function and accept the degraded data
|
|
|
|
// integrity checking
|
|
|
|
return Status::OK();
|
|
|
|
} else {
|
|
|
|
checksum_func =
|
|
|
|
GetCustomChecksumGenerator(requested_checksum_func_name);
|
|
|
|
// we will use the default backup checksum function if the custom
|
|
|
|
// checksum functions is the db standard checksum function but is not
|
|
|
|
// found in the checksum factory passed in; otherwise, we return
|
|
|
|
// Status::InvalidArgument()
|
|
|
|
if (checksum_func == nullptr &&
|
|
|
|
requested_checksum_func_name != kStandardDbFileChecksumFuncName) {
|
|
|
|
return Status::InvalidArgument("Checksum checksum function " +
|
|
|
|
requested_checksum_func_name +
|
|
|
|
" not found");
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
// The requested checksum function is the default backup checksum function
|
|
|
|
return Status::OK();
|
|
|
|
}
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
|
2016-02-29 21:56:55 +01:00
|
|
|
// If size_limit == 0, there is no size limit, copy everything.
|
|
|
|
//
|
|
|
|
// Exactly one of src and contents must be non-empty.
|
|
|
|
//
|
|
|
|
// @param src If non-empty, the file is copied from this pathname.
|
|
|
|
// @param contents If non-empty, the file will be created with these contents.
|
2020-06-25 04:30:15 +02:00
|
|
|
Status CopyOrCreateFile(
|
|
|
|
const std::string& src, const std::string& dst,
|
|
|
|
const std::string& contents, Env* src_env, Env* dst_env,
|
|
|
|
const EnvOptions& src_env_options, bool sync, RateLimiter* rate_limiter,
|
2020-08-12 22:28:43 +02:00
|
|
|
const std::string& backup_checksum_func_name, uint64_t* size = nullptr,
|
|
|
|
std::string* checksum_hex = nullptr,
|
|
|
|
std::string* custom_checksum_hex = nullptr, uint64_t size_limit = 0,
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
std::function<void()> progress_callback = []() {});
|
2014-01-28 18:43:36 +01:00
|
|
|
|
2020-08-12 22:28:43 +02:00
|
|
|
Status CalculateChecksum(
|
|
|
|
const std::string& src, Env* src_env, const EnvOptions& src_env_options,
|
|
|
|
uint64_t size_limit, std::string* checksum_hex,
|
|
|
|
const std::unique_ptr<FileChecksumGenerator>& checksum_func = nullptr,
|
|
|
|
std::string* custom_checksum_hex = nullptr);
|
2014-01-28 18:43:36 +01:00
|
|
|
|
2020-06-25 04:30:15 +02:00
|
|
|
// Obtain db_id and db_session_id from the table properties of file_path
|
2020-07-09 17:36:41 +02:00
|
|
|
Status GetFileDbIdentities(Env* src_env, const EnvOptions& src_env_options,
|
|
|
|
const std::string& file_path, std::string* db_id,
|
|
|
|
std::string* db_session_id);
|
2020-06-25 04:30:15 +02:00
|
|
|
|
2020-08-12 22:28:43 +02:00
|
|
|
Status GetFileChecksumsFromManifestInBackup(Env* src_env,
|
|
|
|
const BackupID& backup_id,
|
|
|
|
const BackupMeta* backup,
|
|
|
|
FileChecksumList* checksum_list);
|
|
|
|
|
|
|
|
Status GetFileChecksumsFromManifest(Env* src_env, const std::string& abs_path,
|
|
|
|
FileChecksumList* checksum_list);
|
|
|
|
|
|
|
|
Status VerifyFileWithCrc32c(Env* src_env, const BackupMeta* backup,
|
|
|
|
const std::string& rel_path);
|
|
|
|
|
|
|
|
struct LogReporter : public log::Reader::Reporter {
|
|
|
|
Status* status;
|
|
|
|
virtual void Corruption(size_t /*bytes*/, const Status& s) override {
|
|
|
|
if (status->ok()) {
|
|
|
|
*status = s;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
};
|
|
|
|
|
2016-02-29 21:56:55 +01:00
|
|
|
struct CopyOrCreateResult {
|
2015-07-02 20:35:51 +02:00
|
|
|
uint64_t size;
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
std::string checksum_hex;
|
2020-08-12 22:28:43 +02:00
|
|
|
std::string custom_checksum_hex;
|
|
|
|
std::string checksum_func_name;
|
2020-06-25 04:30:15 +02:00
|
|
|
std::string db_id;
|
|
|
|
std::string db_session_id;
|
2015-07-02 20:35:51 +02:00
|
|
|
Status status;
|
|
|
|
};
|
2016-02-29 21:56:55 +01:00
|
|
|
|
|
|
|
// Exactly one of src_path and contents must be non-empty. If src_path is
|
|
|
|
// non-empty, the file is copied from this pathname. Otherwise, if contents is
|
|
|
|
// non-empty, the file will be created at dst_path with these contents.
|
|
|
|
struct CopyOrCreateWorkItem {
|
2015-07-02 20:35:51 +02:00
|
|
|
std::string src_path;
|
|
|
|
std::string dst_path;
|
2016-02-29 21:56:55 +01:00
|
|
|
std::string contents;
|
2015-07-02 20:35:51 +02:00
|
|
|
Env* src_env;
|
|
|
|
Env* dst_env;
|
2018-11-13 20:14:41 +01:00
|
|
|
EnvOptions src_env_options;
|
2015-07-02 20:35:51 +02:00
|
|
|
bool sync;
|
2015-09-04 02:00:09 +02:00
|
|
|
RateLimiter* rate_limiter;
|
2015-07-02 20:35:51 +02:00
|
|
|
uint64_t size_limit;
|
2016-02-29 21:56:55 +01:00
|
|
|
std::promise<CopyOrCreateResult> result;
|
2015-09-10 03:49:15 +02:00
|
|
|
std::function<void()> progress_callback;
|
2020-07-03 03:13:31 +02:00
|
|
|
bool verify_checksum_after_work;
|
|
|
|
std::string src_checksum_func_name;
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
std::string src_checksum_hex;
|
2020-08-12 22:28:43 +02:00
|
|
|
std::string backup_checksum_func_name;
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
std::string db_id;
|
|
|
|
std::string db_session_id;
|
2015-07-03 02:23:41 +02:00
|
|
|
|
2017-11-28 23:31:46 +01:00
|
|
|
CopyOrCreateWorkItem()
|
2018-11-13 20:14:41 +01:00
|
|
|
: src_path(""),
|
|
|
|
dst_path(""),
|
|
|
|
contents(""),
|
|
|
|
src_env(nullptr),
|
|
|
|
dst_env(nullptr),
|
|
|
|
src_env_options(),
|
|
|
|
sync(false),
|
|
|
|
rate_limiter(nullptr),
|
2020-07-03 03:13:31 +02:00
|
|
|
size_limit(0),
|
|
|
|
verify_checksum_after_work(false),
|
|
|
|
src_checksum_func_name(kUnknownFileChecksumFuncName),
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
src_checksum_hex(""),
|
2020-08-12 22:28:43 +02:00
|
|
|
backup_checksum_func_name(kUnknownFileChecksumFuncName),
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
db_id(""),
|
|
|
|
db_session_id("") {}
|
2017-11-28 23:31:46 +01:00
|
|
|
|
2016-02-29 21:56:55 +01:00
|
|
|
CopyOrCreateWorkItem(const CopyOrCreateWorkItem&) = delete;
|
|
|
|
CopyOrCreateWorkItem& operator=(const CopyOrCreateWorkItem&) = delete;
|
2015-07-03 02:23:41 +02:00
|
|
|
|
2016-02-29 21:56:55 +01:00
|
|
|
CopyOrCreateWorkItem(CopyOrCreateWorkItem&& o) ROCKSDB_NOEXCEPT {
|
|
|
|
*this = std::move(o);
|
|
|
|
}
|
2015-07-03 02:23:41 +02:00
|
|
|
|
2016-02-29 21:56:55 +01:00
|
|
|
CopyOrCreateWorkItem& operator=(CopyOrCreateWorkItem&& o) ROCKSDB_NOEXCEPT {
|
2015-07-03 02:23:41 +02:00
|
|
|
src_path = std::move(o.src_path);
|
|
|
|
dst_path = std::move(o.dst_path);
|
2016-02-29 21:56:55 +01:00
|
|
|
contents = std::move(o.contents);
|
2015-07-03 02:23:41 +02:00
|
|
|
src_env = o.src_env;
|
|
|
|
dst_env = o.dst_env;
|
2018-11-13 20:14:41 +01:00
|
|
|
src_env_options = std::move(o.src_env_options);
|
2015-07-03 02:23:41 +02:00
|
|
|
sync = o.sync;
|
|
|
|
rate_limiter = o.rate_limiter;
|
|
|
|
size_limit = o.size_limit;
|
|
|
|
result = std::move(o.result);
|
2015-09-10 03:49:15 +02:00
|
|
|
progress_callback = std::move(o.progress_callback);
|
2020-07-03 03:13:31 +02:00
|
|
|
verify_checksum_after_work = o.verify_checksum_after_work;
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
src_checksum_func_name = std::move(o.src_checksum_func_name);
|
|
|
|
src_checksum_hex = std::move(o.src_checksum_hex);
|
2020-08-12 22:28:43 +02:00
|
|
|
backup_checksum_func_name = std::move(o.backup_checksum_func_name);
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
db_id = std::move(o.db_id);
|
|
|
|
db_session_id = std::move(o.db_session_id);
|
2015-07-03 02:23:41 +02:00
|
|
|
return *this;
|
|
|
|
}
|
|
|
|
|
2020-07-03 03:13:31 +02:00
|
|
|
CopyOrCreateWorkItem(
|
|
|
|
std::string _src_path, std::string _dst_path, std::string _contents,
|
|
|
|
Env* _src_env, Env* _dst_env, EnvOptions _src_env_options, bool _sync,
|
|
|
|
RateLimiter* _rate_limiter, uint64_t _size_limit,
|
|
|
|
std::function<void()> _progress_callback = []() {},
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
bool _verify_checksum_after_work = false,
|
2020-07-03 03:13:31 +02:00
|
|
|
const std::string& _src_checksum_func_name =
|
|
|
|
kUnknownFileChecksumFuncName,
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
const std::string& _src_checksum_hex = "",
|
2020-08-12 22:28:43 +02:00
|
|
|
const std::string& _backup_checksum_func_name =
|
|
|
|
kUnknownFileChecksumFuncName,
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
const std::string& _db_id = "", const std::string& _db_session_id = "")
|
2015-07-02 20:35:51 +02:00
|
|
|
: src_path(std::move(_src_path)),
|
|
|
|
dst_path(std::move(_dst_path)),
|
2016-02-29 21:56:55 +01:00
|
|
|
contents(std::move(_contents)),
|
2015-07-02 20:35:51 +02:00
|
|
|
src_env(_src_env),
|
|
|
|
dst_env(_dst_env),
|
2018-11-13 20:14:41 +01:00
|
|
|
src_env_options(std::move(_src_env_options)),
|
2015-07-02 20:35:51 +02:00
|
|
|
sync(_sync),
|
|
|
|
rate_limiter(_rate_limiter),
|
2015-09-10 03:49:15 +02:00
|
|
|
size_limit(_size_limit),
|
2020-07-03 03:13:31 +02:00
|
|
|
progress_callback(_progress_callback),
|
|
|
|
verify_checksum_after_work(_verify_checksum_after_work),
|
|
|
|
src_checksum_func_name(_src_checksum_func_name),
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
src_checksum_hex(_src_checksum_hex),
|
2020-08-12 22:28:43 +02:00
|
|
|
backup_checksum_func_name(_backup_checksum_func_name),
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
db_id(_db_id),
|
|
|
|
db_session_id(_db_session_id) {}
|
2015-07-02 20:35:51 +02:00
|
|
|
};
|
|
|
|
|
2016-02-29 21:56:55 +01:00
|
|
|
struct BackupAfterCopyOrCreateWorkItem {
|
|
|
|
std::future<CopyOrCreateResult> result;
|
2015-07-02 20:35:51 +02:00
|
|
|
bool shared;
|
|
|
|
bool needed_to_copy;
|
|
|
|
Env* backup_env;
|
|
|
|
std::string dst_path_tmp;
|
|
|
|
std::string dst_path;
|
|
|
|
std::string dst_relative;
|
2017-11-28 23:31:46 +01:00
|
|
|
BackupAfterCopyOrCreateWorkItem()
|
|
|
|
: shared(false),
|
|
|
|
needed_to_copy(false),
|
|
|
|
backup_env(nullptr),
|
|
|
|
dst_path_tmp(""),
|
|
|
|
dst_path(""),
|
|
|
|
dst_relative("") {}
|
2015-07-03 02:23:41 +02:00
|
|
|
|
2016-02-29 21:56:55 +01:00
|
|
|
BackupAfterCopyOrCreateWorkItem(BackupAfterCopyOrCreateWorkItem&& o)
|
|
|
|
ROCKSDB_NOEXCEPT {
|
2015-07-03 02:23:41 +02:00
|
|
|
*this = std::move(o);
|
|
|
|
}
|
|
|
|
|
2016-02-29 21:56:55 +01:00
|
|
|
BackupAfterCopyOrCreateWorkItem& operator=(
|
|
|
|
BackupAfterCopyOrCreateWorkItem&& o) ROCKSDB_NOEXCEPT {
|
2015-07-03 02:23:41 +02:00
|
|
|
result = std::move(o.result);
|
|
|
|
shared = o.shared;
|
|
|
|
needed_to_copy = o.needed_to_copy;
|
|
|
|
backup_env = o.backup_env;
|
|
|
|
dst_path_tmp = std::move(o.dst_path_tmp);
|
|
|
|
dst_path = std::move(o.dst_path);
|
|
|
|
dst_relative = std::move(o.dst_relative);
|
|
|
|
return *this;
|
|
|
|
}
|
|
|
|
|
2016-02-29 21:56:55 +01:00
|
|
|
BackupAfterCopyOrCreateWorkItem(std::future<CopyOrCreateResult>&& _result,
|
|
|
|
bool _shared, bool _needed_to_copy,
|
|
|
|
Env* _backup_env, std::string _dst_path_tmp,
|
|
|
|
std::string _dst_path,
|
|
|
|
std::string _dst_relative)
|
2015-07-02 20:35:51 +02:00
|
|
|
: result(std::move(_result)),
|
|
|
|
shared(_shared),
|
|
|
|
needed_to_copy(_needed_to_copy),
|
|
|
|
backup_env(_backup_env),
|
|
|
|
dst_path_tmp(std::move(_dst_path_tmp)),
|
|
|
|
dst_path(std::move(_dst_path)),
|
|
|
|
dst_relative(std::move(_dst_relative)) {}
|
|
|
|
};
|
|
|
|
|
2016-02-29 21:56:55 +01:00
|
|
|
struct RestoreAfterCopyOrCreateWorkItem {
|
|
|
|
std::future<CopyOrCreateResult> result;
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
std::string checksum_hex;
|
|
|
|
RestoreAfterCopyOrCreateWorkItem() : checksum_hex("") {}
|
2016-02-29 21:56:55 +01:00
|
|
|
RestoreAfterCopyOrCreateWorkItem(std::future<CopyOrCreateResult>&& _result,
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
const std::string& _checksum_hex)
|
|
|
|
: result(std::move(_result)), checksum_hex(_checksum_hex) {}
|
2016-02-29 21:56:55 +01:00
|
|
|
RestoreAfterCopyOrCreateWorkItem(RestoreAfterCopyOrCreateWorkItem&& o)
|
|
|
|
ROCKSDB_NOEXCEPT {
|
2015-07-03 02:23:41 +02:00
|
|
|
*this = std::move(o);
|
|
|
|
}
|
|
|
|
|
2016-02-29 21:56:55 +01:00
|
|
|
RestoreAfterCopyOrCreateWorkItem& operator=(
|
|
|
|
RestoreAfterCopyOrCreateWorkItem&& o) ROCKSDB_NOEXCEPT {
|
2015-07-03 02:23:41 +02:00
|
|
|
result = std::move(o.result);
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
checksum_hex = std::move(o.checksum_hex);
|
2015-07-03 02:23:41 +02:00
|
|
|
return *this;
|
|
|
|
}
|
2015-07-02 20:35:51 +02:00
|
|
|
};
|
|
|
|
|
2015-07-14 20:51:36 +02:00
|
|
|
bool initialized_;
|
2015-09-10 03:49:15 +02:00
|
|
|
std::mutex byte_report_mutex_;
|
2016-02-29 21:56:55 +01:00
|
|
|
channel<CopyOrCreateWorkItem> files_to_copy_or_create_;
|
2017-02-06 23:43:55 +01:00
|
|
|
std::vector<port::Thread> threads_;
|
2020-03-29 04:05:54 +02:00
|
|
|
std::atomic<CpuPriority> threads_cpu_priority_;
|
Auto-GarbageCollect on PurgeOldBackups and DeleteBackup (#6015)
Summary:
Only if there is a crash, power failure, or I/O error in
DeleteBackup, shared or private files from the backup might be left
behind that are not cleaned up by PurgeOldBackups or DeleteBackup-- only
by GarbageCollect. This makes the BackupEngine API "leaky by default."
Even if it means a modest performance hit, I think we should make
Delete and Purge do as they say, with ongoing best effort: i.e. future
calls will attempt to finish any incomplete work from earlier calls.
This change does that by having DeleteBackup and PurgeOldBackups do a
GarbageCollect, unless (to minimize performance hit) this BackupEngine
has already done a GarbageCollect and there have been no
deletion-related I/O errors in that GarbageCollect or since then.
Rejected alternative 1: remove meta file last instead of first. This would in theory turn partially deleted backups into corrupted backups, but code changes would be needed to allow the missing files and consider it acceptably corrupt, rather than failing to open the BackupEngine. This might be a reasonable choice, but I mostly rejected it because it doesn't solve the legacy problem of cleaning up existing lingering files.
Rejected alternative 2: use a deletion marker file. If deletion started with creating a file that marks a backup as flagged for deletion, then we could reliably detect partially deleted backups and efficiently finish removing them. In addition to not solving the legacy problem, this could be precarious if there's a disk full situation, and we try to create a new file in order to delete some files. Ugh.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6015
Test Plan: Updated unit tests
Differential Revision: D18401333
Pulled By: pdillinger
fbshipit-source-id: 12944e372ce6809f3f5a4c416c3b321a8927d925
2019-11-09 04:13:41 +01:00
|
|
|
// Certain operations like PurgeOldBackups and DeleteBackup will trigger
|
|
|
|
// automatic GarbageCollect (true) unless we've already done one in this
|
|
|
|
// session and have not failed to delete backup files since then (false).
|
|
|
|
bool might_need_garbage_collect_ = true;
|
2015-07-02 20:35:51 +02:00
|
|
|
|
2016-02-29 21:56:55 +01:00
|
|
|
// Adds a file to the backup work queue to be copied or created if it doesn't
|
|
|
|
// already exist.
|
|
|
|
//
|
|
|
|
// Exactly one of src_dir and contents must be non-empty.
|
|
|
|
//
|
|
|
|
// @param src_dir If non-empty, the file in this directory named fname will be
|
|
|
|
// copied.
|
|
|
|
// @param fname Name of destination file and, in case of copy, source file.
|
|
|
|
// @param contents If non-empty, the file will be created with these contents.
|
2015-07-02 20:35:51 +02:00
|
|
|
Status AddBackupFileWorkItem(
|
2015-09-10 03:49:15 +02:00
|
|
|
std::unordered_set<std::string>& live_dst_paths,
|
2016-02-29 21:56:55 +01:00
|
|
|
std::vector<BackupAfterCopyOrCreateWorkItem>& backup_items_to_finish,
|
2015-09-10 03:49:15 +02:00
|
|
|
BackupID backup_id, bool shared, const std::string& src_dir,
|
2016-02-29 21:56:55 +01:00
|
|
|
const std::string& fname, // starts with "/"
|
2018-11-13 20:14:41 +01:00
|
|
|
const EnvOptions& src_env_options, RateLimiter* rate_limiter,
|
|
|
|
uint64_t size_bytes, uint64_t size_limit = 0,
|
2015-09-10 03:49:15 +02:00
|
|
|
bool shared_checksum = false,
|
2016-02-29 21:56:55 +01:00
|
|
|
std::function<void()> progress_callback = []() {},
|
2020-07-03 03:13:31 +02:00
|
|
|
const std::string& contents = std::string(),
|
|
|
|
const std::string& src_checksum_func_name = kUnknownFileChecksumFuncName,
|
|
|
|
const std::string& src_checksum_str = kUnknownFileChecksum);
|
2015-07-02 20:35:51 +02:00
|
|
|
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
// backup state data
|
|
|
|
BackupID latest_backup_id_;
|
2017-09-01 00:30:35 +02:00
|
|
|
BackupID latest_valid_backup_id_;
|
2018-11-09 20:17:34 +01:00
|
|
|
std::map<BackupID, std::unique_ptr<BackupMeta>> backups_;
|
|
|
|
std::map<BackupID, std::pair<Status, std::unique_ptr<BackupMeta>>>
|
|
|
|
corrupt_backups_;
|
2015-01-05 14:35:09 +01:00
|
|
|
std::unordered_map<std::string,
|
|
|
|
std::shared_ptr<FileInfo>> backuped_file_infos_;
|
2014-01-09 21:24:28 +01:00
|
|
|
std::atomic<bool> stop_backup_;
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
|
|
|
|
// options data
|
|
|
|
BackupableDBOptions options_;
|
|
|
|
Env* db_env_;
|
|
|
|
Env* backup_env_;
|
|
|
|
|
2014-03-05 02:02:25 +01:00
|
|
|
// directories
|
2018-11-09 20:17:34 +01:00
|
|
|
std::unique_ptr<Directory> backup_directory_;
|
|
|
|
std::unique_ptr<Directory> shared_directory_;
|
|
|
|
std::unique_ptr<Directory> meta_directory_;
|
|
|
|
std::unique_ptr<Directory> private_directory_;
|
2014-03-05 02:02:25 +01:00
|
|
|
|
2014-03-24 19:38:44 +01:00
|
|
|
static const size_t kDefaultCopyFileBufferSize = 5 * 1024 * 1024LL; // 5MB
|
|
|
|
size_t copy_file_buffer_size_;
|
2014-04-25 21:49:29 +02:00
|
|
|
bool read_only_;
|
2014-09-09 22:44:42 +02:00
|
|
|
BackupStatistics backup_statistics_;
|
2016-04-01 19:56:52 +02:00
|
|
|
static const size_t kMaxAppMetaSize = 1024 * 1024; // 1MB
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
};
|
|
|
|
|
2020-03-29 04:05:54 +02:00
|
|
|
Status BackupEngine::Open(const BackupableDBOptions& options, Env* env,
|
2014-11-13 18:51:41 +01:00
|
|
|
BackupEngine** backup_engine_ptr) {
|
2015-07-14 20:51:36 +02:00
|
|
|
std::unique_ptr<BackupEngineImpl> backup_engine(
|
2020-03-29 04:05:54 +02:00
|
|
|
new BackupEngineImpl(options, env));
|
2015-07-14 20:51:36 +02:00
|
|
|
auto s = backup_engine->Initialize();
|
|
|
|
if (!s.ok()) {
|
|
|
|
*backup_engine_ptr = nullptr;
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
*backup_engine_ptr = backup_engine.release();
|
2014-11-13 18:51:41 +01:00
|
|
|
return Status::OK();
|
|
|
|
}
|
|
|
|
|
2020-03-29 04:05:54 +02:00
|
|
|
BackupEngineImpl::BackupEngineImpl(const BackupableDBOptions& options,
|
|
|
|
Env* db_env, bool read_only)
|
2015-07-14 20:51:36 +02:00
|
|
|
: initialized_(false),
|
2020-06-05 00:32:29 +02:00
|
|
|
threads_cpu_priority_(),
|
2017-11-28 23:31:46 +01:00
|
|
|
latest_backup_id_(0),
|
|
|
|
latest_valid_backup_id_(0),
|
2015-07-14 20:51:36 +02:00
|
|
|
stop_backup_(false),
|
2014-01-09 21:24:28 +01:00
|
|
|
options_(options),
|
|
|
|
db_env_(db_env),
|
2014-03-24 19:38:44 +01:00
|
|
|
backup_env_(options.backup_env != nullptr ? options.backup_env : db_env_),
|
2014-04-25 21:49:29 +02:00
|
|
|
copy_file_buffer_size_(kDefaultCopyFileBufferSize),
|
2016-06-04 02:02:07 +02:00
|
|
|
read_only_(read_only) {
|
|
|
|
if (options_.backup_rate_limiter == nullptr &&
|
|
|
|
options_.backup_rate_limit > 0) {
|
|
|
|
options_.backup_rate_limiter.reset(
|
|
|
|
NewGenericRateLimiter(options_.backup_rate_limit));
|
|
|
|
}
|
|
|
|
if (options_.restore_rate_limiter == nullptr &&
|
|
|
|
options_.restore_rate_limit > 0) {
|
|
|
|
options_.restore_rate_limiter.reset(
|
|
|
|
NewGenericRateLimiter(options_.restore_rate_limit));
|
|
|
|
}
|
|
|
|
}
|
2015-07-14 20:51:36 +02:00
|
|
|
|
|
|
|
BackupEngineImpl::~BackupEngineImpl() {
|
2016-02-29 21:56:55 +01:00
|
|
|
files_to_copy_or_create_.sendEof();
|
2015-07-14 20:51:36 +02:00
|
|
|
for (auto& t : threads_) {
|
|
|
|
t.join();
|
|
|
|
}
|
|
|
|
LogFlush(options_.info_log);
|
|
|
|
}
|
2015-07-02 20:35:51 +02:00
|
|
|
|
2015-07-14 20:51:36 +02:00
|
|
|
Status BackupEngineImpl::Initialize() {
|
|
|
|
assert(!initialized_);
|
|
|
|
initialized_ = true;
|
2014-04-25 21:49:29 +02:00
|
|
|
if (read_only_) {
|
2017-03-16 03:22:52 +01:00
|
|
|
ROCKS_LOG_INFO(options_.info_log, "Starting read_only backup engine");
|
2014-04-25 21:49:29 +02:00
|
|
|
}
|
2014-03-10 20:09:54 +01:00
|
|
|
options_.Dump(options_.info_log);
|
|
|
|
|
2014-04-25 21:49:29 +02:00
|
|
|
if (!read_only_) {
|
Auto-GarbageCollect on PurgeOldBackups and DeleteBackup (#6015)
Summary:
Only if there is a crash, power failure, or I/O error in
DeleteBackup, shared or private files from the backup might be left
behind that are not cleaned up by PurgeOldBackups or DeleteBackup-- only
by GarbageCollect. This makes the BackupEngine API "leaky by default."
Even if it means a modest performance hit, I think we should make
Delete and Purge do as they say, with ongoing best effort: i.e. future
calls will attempt to finish any incomplete work from earlier calls.
This change does that by having DeleteBackup and PurgeOldBackups do a
GarbageCollect, unless (to minimize performance hit) this BackupEngine
has already done a GarbageCollect and there have been no
deletion-related I/O errors in that GarbageCollect or since then.
Rejected alternative 1: remove meta file last instead of first. This would in theory turn partially deleted backups into corrupted backups, but code changes would be needed to allow the missing files and consider it acceptably corrupt, rather than failing to open the BackupEngine. This might be a reasonable choice, but I mostly rejected it because it doesn't solve the legacy problem of cleaning up existing lingering files.
Rejected alternative 2: use a deletion marker file. If deletion started with creating a file that marks a backup as flagged for deletion, then we could reliably detect partially deleted backups and efficiently finish removing them. In addition to not solving the legacy problem, this could be precarious if there's a disk full situation, and we try to create a new file in order to delete some files. Ugh.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6015
Test Plan: Updated unit tests
Differential Revision: D18401333
Pulled By: pdillinger
fbshipit-source-id: 12944e372ce6809f3f5a4c416c3b321a8927d925
2019-11-09 04:13:41 +01:00
|
|
|
// we might need to clean up from previous crash or I/O errors
|
|
|
|
might_need_garbage_collect_ = true;
|
|
|
|
|
2019-11-25 23:18:10 +01:00
|
|
|
if (options_.max_valid_backups_to_open != port::kMaxInt32) {
|
|
|
|
options_.max_valid_backups_to_open = port::kMaxInt32;
|
|
|
|
ROCKS_LOG_WARN(
|
|
|
|
options_.info_log,
|
|
|
|
"`max_valid_backups_to_open` is not set to the default value. Ignoring "
|
|
|
|
"its value since BackupEngine is not read-only.");
|
|
|
|
}
|
|
|
|
|
2015-07-14 20:51:36 +02:00
|
|
|
// gather the list of directories that we need to create
|
|
|
|
std::vector<std::pair<std::string, std::unique_ptr<Directory>*>>
|
|
|
|
directories;
|
|
|
|
directories.emplace_back(GetAbsolutePath(), &backup_directory_);
|
2014-04-25 21:49:29 +02:00
|
|
|
if (options_.share_table_files) {
|
2014-05-03 02:08:55 +02:00
|
|
|
if (options_.share_files_with_checksum) {
|
2015-07-14 20:51:36 +02:00
|
|
|
directories.emplace_back(
|
|
|
|
GetAbsolutePath(GetSharedFileWithChecksumRel()),
|
|
|
|
&shared_directory_);
|
2014-05-03 02:08:55 +02:00
|
|
|
} else {
|
2015-07-14 20:51:36 +02:00
|
|
|
directories.emplace_back(GetAbsolutePath(GetSharedFileRel()),
|
|
|
|
&shared_directory_);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
directories.emplace_back(GetAbsolutePath(GetPrivateDirRel()),
|
|
|
|
&private_directory_);
|
|
|
|
directories.emplace_back(GetBackupMetaDir(), &meta_directory_);
|
|
|
|
// create all the dirs we need
|
|
|
|
for (const auto& d : directories) {
|
|
|
|
auto s = backup_env_->CreateDirIfMissing(d.first);
|
|
|
|
if (s.ok()) {
|
|
|
|
s = backup_env_->NewDirectory(d.first, d.second);
|
|
|
|
}
|
|
|
|
if (!s.ok()) {
|
|
|
|
return s;
|
2014-05-03 02:08:55 +02:00
|
|
|
}
|
2014-04-25 21:49:29 +02:00
|
|
|
}
|
2014-01-09 21:24:28 +01:00
|
|
|
}
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
|
|
|
|
std::vector<std::string> backup_meta_files;
|
2015-07-14 20:51:36 +02:00
|
|
|
{
|
|
|
|
auto s = backup_env_->GetChildren(GetBackupMetaDir(), &backup_meta_files);
|
2016-12-12 22:18:38 +01:00
|
|
|
if (s.IsNotFound()) {
|
2016-07-02 00:28:08 +02:00
|
|
|
return Status::NotFound(GetBackupMetaDir() + " is missing");
|
2016-12-12 22:18:38 +01:00
|
|
|
} else if (!s.ok()) {
|
|
|
|
return s;
|
2015-07-14 20:51:36 +02:00
|
|
|
}
|
|
|
|
}
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
// create backups_ structure
|
|
|
|
for (auto& file : backup_meta_files) {
|
2015-02-27 23:03:56 +01:00
|
|
|
if (file == "." || file == "..") {
|
|
|
|
continue;
|
|
|
|
}
|
2017-03-16 03:22:52 +01:00
|
|
|
ROCKS_LOG_INFO(options_.info_log, "Detected backup %s", file.c_str());
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
BackupID backup_id = 0;
|
|
|
|
sscanf(file.c_str(), "%u", &backup_id);
|
2020-02-20 21:07:53 +01:00
|
|
|
if (backup_id == 0 || file != ROCKSDB_NAMESPACE::ToString(backup_id)) {
|
2014-04-25 21:49:29 +02:00
|
|
|
if (!read_only_) {
|
|
|
|
// invalid file name, delete that
|
2015-07-14 20:51:36 +02:00
|
|
|
auto s = backup_env_->DeleteFile(GetBackupMetaDir() + "/" + file);
|
2017-03-16 03:22:52 +01:00
|
|
|
ROCKS_LOG_INFO(options_.info_log,
|
|
|
|
"Unrecognized meta file %s, deleting -- %s",
|
|
|
|
file.c_str(), s.ToString().c_str());
|
2014-04-25 21:49:29 +02:00
|
|
|
}
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
continue;
|
|
|
|
}
|
|
|
|
assert(backups_.find(backup_id) == backups_.end());
|
2020-06-25 04:30:15 +02:00
|
|
|
// Insert all the (backup_id, BackupMeta) that will be loaded later
|
|
|
|
// The loading performed later will check whether there are corrupt backups
|
|
|
|
// and move the corrupt backups to corrupt_backups_
|
2018-02-22 02:33:14 +01:00
|
|
|
backups_.insert(std::make_pair(
|
2018-11-09 20:17:34 +01:00
|
|
|
backup_id, std::unique_ptr<BackupMeta>(new BackupMeta(
|
2018-02-22 02:33:14 +01:00
|
|
|
GetBackupMetaFile(backup_id, false /* tmp */),
|
|
|
|
GetBackupMetaFile(backup_id, true /* tmp */),
|
|
|
|
&backuped_file_infos_, backup_env_))));
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
}
|
|
|
|
|
Remove the need for LATEST_BACKUP in BackupEngine
Summary:
In the first implementation of BackupEngine, LATEST_BACKUP was the commit point. The backup became committed after the write to LATEST_BACKUP completed.
However, we can avoid the need for LATEST_BACKUP. Instead of write to LATEST_BACKUP, the commit point can be the rename from `meta/<backup_id>.tmp` to `meta/<backup_id>`. Once we see that there exists a file `meta/<backup_id>` (without tmp), we can assume that backup is valid.
In this diff, we still write out the file LATEST_BACKUP. We need to do this so that we can maintain backward compatibility. However, the new version doesn't depend on this file anymore. We get the latest backup by `ls`-ing `meta` directory.
This diff depends on D41925
Test Plan: Adjusted backupable_db_test to this new behavior
Reviewers: benj, yhchiang, sdong, AaronFeldman
Reviewed By: sdong
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D42069
2015-09-02 20:49:49 +02:00
|
|
|
latest_backup_id_ = 0;
|
2017-09-01 00:30:35 +02:00
|
|
|
latest_valid_backup_id_ = 0;
|
2015-04-25 11:14:27 +02:00
|
|
|
if (options_.destroy_old_data) { // Destroy old data
|
2014-04-25 21:49:29 +02:00
|
|
|
assert(!read_only_);
|
2017-03-16 03:22:52 +01:00
|
|
|
ROCKS_LOG_INFO(
|
|
|
|
options_.info_log,
|
2015-02-27 23:03:56 +01:00
|
|
|
"Backup Engine started with destroy_old_data == true, deleting all "
|
|
|
|
"backups");
|
2015-07-14 20:51:36 +02:00
|
|
|
auto s = PurgeOldBackups(0);
|
|
|
|
if (s.ok()) {
|
|
|
|
s = GarbageCollect();
|
|
|
|
}
|
|
|
|
if (!s.ok()) {
|
|
|
|
return s;
|
|
|
|
}
|
2014-05-03 02:08:55 +02:00
|
|
|
} else { // Load data from storage
|
2020-06-25 04:30:15 +02:00
|
|
|
// abs_path_to_size: maps absolute paths of files in backup directory to
|
|
|
|
// their corresponding sizes
|
2016-03-02 04:33:33 +01:00
|
|
|
std::unordered_map<std::string, uint64_t> abs_path_to_size;
|
2020-06-25 04:30:15 +02:00
|
|
|
// Insert files and their sizes in backup sub-directories (shared and
|
|
|
|
// shared_checksum) to abs_path_to_size
|
2016-03-02 04:33:33 +01:00
|
|
|
for (const auto& rel_dir :
|
|
|
|
{GetSharedFileRel(), GetSharedFileWithChecksumRel()}) {
|
|
|
|
const auto abs_dir = GetAbsolutePath(rel_dir);
|
2016-04-26 21:33:30 +02:00
|
|
|
InsertPathnameToSizeBytes(abs_dir, backup_env_, &abs_path_to_size);
|
2016-03-02 04:33:33 +01:00
|
|
|
}
|
2017-04-19 22:15:16 +02:00
|
|
|
// load the backups if any, until valid_backups_to_open of the latest
|
|
|
|
// non-corrupted backups have been successfully opened.
|
2017-09-12 22:12:58 +02:00
|
|
|
int valid_backups_to_open = options_.max_valid_backups_to_open;
|
2017-04-19 22:15:16 +02:00
|
|
|
for (auto backup_iter = backups_.rbegin();
|
2017-09-12 22:12:58 +02:00
|
|
|
backup_iter != backups_.rend();
|
2017-04-19 22:15:16 +02:00
|
|
|
++backup_iter) {
|
2017-09-01 00:30:35 +02:00
|
|
|
assert(latest_backup_id_ == 0 || latest_backup_id_ > backup_iter->first);
|
|
|
|
if (latest_backup_id_ == 0) {
|
|
|
|
latest_backup_id_ = backup_iter->first;
|
|
|
|
}
|
2017-09-12 22:12:58 +02:00
|
|
|
if (valid_backups_to_open == 0) {
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
2020-06-25 04:30:15 +02:00
|
|
|
// Insert files and their sizes in backup sub-directories
|
|
|
|
// (private/backup_id) to abs_path_to_size
|
2016-03-02 04:33:33 +01:00
|
|
|
InsertPathnameToSizeBytes(
|
2017-04-19 22:15:16 +02:00
|
|
|
GetAbsolutePath(GetPrivateFileRel(backup_iter->first)), backup_env_,
|
2016-04-26 21:33:30 +02:00
|
|
|
&abs_path_to_size);
|
2017-04-19 22:15:16 +02:00
|
|
|
Status s = backup_iter->second->LoadFromFile(options_.backup_dir,
|
|
|
|
abs_path_to_size);
|
2016-12-15 01:29:23 +01:00
|
|
|
if (s.IsCorruption()) {
|
2017-03-16 03:22:52 +01:00
|
|
|
ROCKS_LOG_INFO(options_.info_log, "Backup %u corrupted -- %s",
|
2017-04-19 22:15:16 +02:00
|
|
|
backup_iter->first, s.ToString().c_str());
|
|
|
|
corrupt_backups_.insert(
|
|
|
|
std::make_pair(backup_iter->first,
|
|
|
|
std::make_pair(s, std::move(backup_iter->second))));
|
2016-12-15 01:29:23 +01:00
|
|
|
} else if (!s.ok()) {
|
|
|
|
// Distinguish corruption errors from errors in the backup Env.
|
|
|
|
// Errors in the backup Env (i.e., this code path) will cause Open() to
|
|
|
|
// fail, whereas corruption errors would not cause Open() failures.
|
|
|
|
return s;
|
2015-02-27 23:03:56 +01:00
|
|
|
} else {
|
2017-03-16 03:22:52 +01:00
|
|
|
ROCKS_LOG_INFO(options_.info_log, "Loading backup %" PRIu32 " OK:\n%s",
|
2017-04-19 22:15:16 +02:00
|
|
|
backup_iter->first,
|
|
|
|
backup_iter->second->GetInfoString().c_str());
|
2017-09-01 00:30:35 +02:00
|
|
|
assert(latest_valid_backup_id_ == 0 ||
|
|
|
|
latest_valid_backup_id_ > backup_iter->first);
|
|
|
|
if (latest_valid_backup_id_ == 0) {
|
|
|
|
latest_valid_backup_id_ = backup_iter->first;
|
|
|
|
}
|
2017-04-19 22:15:16 +02:00
|
|
|
--valid_backups_to_open;
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
}
|
|
|
|
}
|
2014-11-13 18:51:41 +01:00
|
|
|
|
2015-01-05 14:35:09 +01:00
|
|
|
for (const auto& corrupt : corrupt_backups_) {
|
2014-11-13 18:51:41 +01:00
|
|
|
backups_.erase(backups_.find(corrupt.first));
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
}
|
2017-04-19 22:15:16 +02:00
|
|
|
// erase the backups before max_valid_backups_to_open
|
|
|
|
int num_unopened_backups;
|
|
|
|
if (options_.max_valid_backups_to_open == 0) {
|
|
|
|
num_unopened_backups = 0;
|
|
|
|
} else {
|
|
|
|
num_unopened_backups =
|
|
|
|
std::max(0, static_cast<int>(backups_.size()) -
|
|
|
|
options_.max_valid_backups_to_open);
|
|
|
|
}
|
|
|
|
for (int i = 0; i < num_unopened_backups; ++i) {
|
|
|
|
assert(backups_.begin()->second->Empty());
|
|
|
|
backups_.erase(backups_.begin());
|
|
|
|
}
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
}
|
|
|
|
|
2017-03-16 03:22:52 +01:00
|
|
|
ROCKS_LOG_INFO(options_.info_log, "Latest backup is %u", latest_backup_id_);
|
2017-09-01 00:30:35 +02:00
|
|
|
ROCKS_LOG_INFO(options_.info_log, "Latest valid backup is %u",
|
|
|
|
latest_valid_backup_id_);
|
2015-02-27 23:03:56 +01:00
|
|
|
|
2016-02-29 21:56:55 +01:00
|
|
|
// set up threads perform copies from files_to_copy_or_create_ in the
|
|
|
|
// background
|
2020-03-29 04:05:54 +02:00
|
|
|
threads_cpu_priority_ = CpuPriority::kNormal;
|
|
|
|
threads_.reserve(options_.max_background_operations);
|
2015-07-07 19:49:16 +02:00
|
|
|
for (int t = 0; t < options_.max_background_operations; t++) {
|
2016-05-20 03:57:47 +02:00
|
|
|
threads_.emplace_back([this]() {
|
2017-03-30 01:58:45 +02:00
|
|
|
#if defined(_GNU_SOURCE) && defined(__GLIBC_PREREQ)
|
|
|
|
#if __GLIBC_PREREQ(2, 12)
|
|
|
|
pthread_setname_np(pthread_self(), "backup_engine");
|
|
|
|
#endif
|
|
|
|
#endif
|
2020-03-29 04:05:54 +02:00
|
|
|
CpuPriority current_priority = CpuPriority::kNormal;
|
2016-02-29 21:56:55 +01:00
|
|
|
CopyOrCreateWorkItem work_item;
|
|
|
|
while (files_to_copy_or_create_.read(work_item)) {
|
2020-03-29 04:05:54 +02:00
|
|
|
CpuPriority priority = threads_cpu_priority_;
|
|
|
|
if (current_priority != priority) {
|
|
|
|
TEST_SYNC_POINT_CALLBACK(
|
|
|
|
"BackupEngineImpl::Initialize:SetCpuPriority", &priority);
|
|
|
|
port::SetCpuPriority(0, priority);
|
|
|
|
current_priority = priority;
|
|
|
|
}
|
2016-02-29 21:56:55 +01:00
|
|
|
CopyOrCreateResult result;
|
|
|
|
result.status = CopyOrCreateFile(
|
|
|
|
work_item.src_path, work_item.dst_path, work_item.contents,
|
2018-11-13 20:14:41 +01:00
|
|
|
work_item.src_env, work_item.dst_env, work_item.src_env_options,
|
2020-08-12 22:28:43 +02:00
|
|
|
work_item.sync, work_item.rate_limiter,
|
|
|
|
work_item.backup_checksum_func_name, &result.size,
|
|
|
|
&result.checksum_hex, &result.custom_checksum_hex,
|
|
|
|
work_item.size_limit, work_item.progress_callback);
|
|
|
|
result.checksum_func_name = work_item.backup_checksum_func_name;
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
result.db_id = work_item.db_id;
|
|
|
|
result.db_session_id = work_item.db_session_id;
|
2020-07-03 03:13:31 +02:00
|
|
|
if (result.status.ok() && work_item.verify_checksum_after_work) {
|
2020-08-12 22:28:43 +02:00
|
|
|
// work_item.verify_checksum_after_work being true means backup engine
|
|
|
|
// has obtained its crc32c and/or custom checksum for the table file.
|
|
|
|
// Therefore, we can try to compare the checksums if possible.
|
2020-07-03 03:13:31 +02:00
|
|
|
if (work_item.src_checksum_func_name ==
|
|
|
|
kUnknownFileChecksumFuncName ||
|
2020-08-12 22:28:43 +02:00
|
|
|
IsSameChecksumFunc(result.checksum_func_name,
|
|
|
|
work_item.src_checksum_func_name)) {
|
|
|
|
std::string checksum_to_compare;
|
|
|
|
std::string checksum_func_name_used;
|
|
|
|
if (work_item.src_checksum_func_name ==
|
|
|
|
kUnknownFileChecksumFuncName ||
|
|
|
|
work_item.src_checksum_func_name ==
|
|
|
|
kStandardDbFileChecksumFuncName) {
|
|
|
|
// kUnknownFileChecksumFuncName implies no table file checksums in
|
|
|
|
// db manifest, but we can compare using the crc32c checksum
|
|
|
|
checksum_to_compare = result.checksum_hex;
|
|
|
|
checksum_func_name_used = kStandardDbFileChecksumFuncName;
|
|
|
|
} else {
|
|
|
|
checksum_to_compare = result.custom_checksum_hex;
|
|
|
|
checksum_func_name_used = work_item.src_checksum_func_name;
|
|
|
|
}
|
|
|
|
if (work_item.src_checksum_hex != checksum_to_compare) {
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
std::string checksum_info(
|
|
|
|
"Expected checksum is " + work_item.src_checksum_hex +
|
2020-08-12 22:28:43 +02:00
|
|
|
" while computed checksum is " + checksum_to_compare);
|
|
|
|
result.status = Status::Corruption(
|
|
|
|
checksum_func_name_used + " mismatch after copying to " +
|
|
|
|
work_item.dst_path + ": " + checksum_info);
|
2020-07-03 03:13:31 +02:00
|
|
|
}
|
|
|
|
} else {
|
|
|
|
std::string checksum_function_info(
|
|
|
|
"Existing checksum function is " +
|
|
|
|
work_item.src_checksum_func_name +
|
|
|
|
" while provided checksum function is " +
|
2020-08-12 22:28:43 +02:00
|
|
|
result.checksum_func_name);
|
2020-07-03 03:13:31 +02:00
|
|
|
ROCKS_LOG_INFO(
|
|
|
|
options_.info_log,
|
|
|
|
"Unable to verify checksum after copying to %s: %s\n",
|
|
|
|
work_item.dst_path.c_str(), checksum_function_info.c_str());
|
|
|
|
}
|
|
|
|
}
|
2015-07-07 19:49:16 +02:00
|
|
|
work_item.result.set_value(std::move(result));
|
|
|
|
}
|
|
|
|
});
|
|
|
|
}
|
2017-03-16 03:22:52 +01:00
|
|
|
ROCKS_LOG_INFO(options_.info_log, "Initialized BackupEngine");
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
|
2015-07-14 20:51:36 +02:00
|
|
|
return Status::OK();
|
2015-07-02 20:35:51 +02:00
|
|
|
}
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
|
2016-04-01 19:56:52 +02:00
|
|
|
Status BackupEngineImpl::CreateNewBackupWithMetadata(
|
2020-03-29 04:05:54 +02:00
|
|
|
const CreateBackupOptions& options, DB* db,
|
|
|
|
const std::string& app_metadata) {
|
2015-07-14 20:51:36 +02:00
|
|
|
assert(initialized_);
|
2014-04-25 21:49:29 +02:00
|
|
|
assert(!read_only_);
|
2016-04-01 19:56:52 +02:00
|
|
|
if (app_metadata.size() > kMaxAppMetaSize) {
|
|
|
|
return Status::InvalidArgument("App metadata too large");
|
|
|
|
}
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
|
2020-03-29 04:05:54 +02:00
|
|
|
if (options.decrease_background_thread_cpu_priority) {
|
|
|
|
if (options.background_thread_cpu_priority < threads_cpu_priority_) {
|
|
|
|
threads_cpu_priority_.store(options.background_thread_cpu_priority);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
BackupID new_backup_id = latest_backup_id_ + 1;
|
2016-06-10 04:03:10 +02:00
|
|
|
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
assert(backups_.find(new_backup_id) == backups_.end());
|
2018-04-21 02:23:34 +02:00
|
|
|
|
|
|
|
auto private_dir = GetAbsolutePath(GetPrivateFileRel(new_backup_id));
|
|
|
|
Status s = backup_env_->FileExists(private_dir);
|
|
|
|
if (s.ok()) {
|
|
|
|
// maybe last backup failed and left partial state behind, clean it up.
|
|
|
|
// need to do this before updating backups_ such that a private dir
|
2019-11-14 15:18:23 +01:00
|
|
|
// named after new_backup_id will be cleaned up.
|
|
|
|
// (If an incomplete new backup is followed by an incomplete delete
|
|
|
|
// of the latest full backup, then there could be more than one next
|
|
|
|
// id with a private dir, the last thing to be deleted in delete
|
|
|
|
// backup, but all will be cleaned up with a GarbageCollect.)
|
2018-04-21 02:23:34 +02:00
|
|
|
s = GarbageCollect();
|
|
|
|
} else if (s.IsNotFound()) {
|
|
|
|
// normal case, the new backup's private dir doesn't exist yet
|
|
|
|
s = Status::OK();
|
|
|
|
}
|
|
|
|
|
2018-02-22 02:33:14 +01:00
|
|
|
auto ret = backups_.insert(std::make_pair(
|
2018-11-09 20:17:34 +01:00
|
|
|
new_backup_id, std::unique_ptr<BackupMeta>(new BackupMeta(
|
2018-02-22 02:33:14 +01:00
|
|
|
GetBackupMetaFile(new_backup_id, false /* tmp */),
|
|
|
|
GetBackupMetaFile(new_backup_id, true /* tmp */),
|
|
|
|
&backuped_file_infos_, backup_env_))));
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
assert(ret.second == true);
|
|
|
|
auto& new_backup = ret.first->second;
|
2015-01-05 14:35:09 +01:00
|
|
|
new_backup->RecordTimestamp();
|
2016-04-01 19:56:52 +02:00
|
|
|
new_backup->SetAppMetadata(app_metadata);
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
|
2018-04-21 02:23:34 +02:00
|
|
|
auto start_backup = backup_env_->NowMicros();
|
2014-09-09 22:44:42 +02:00
|
|
|
|
2017-03-16 03:22:52 +01:00
|
|
|
ROCKS_LOG_INFO(options_.info_log,
|
|
|
|
"Started the backup process -- creating backup %u",
|
|
|
|
new_backup_id);
|
2017-02-23 02:34:12 +01:00
|
|
|
if (s.ok()) {
|
2018-04-21 02:23:34 +02:00
|
|
|
s = backup_env_->CreateDir(private_dir);
|
2017-02-23 02:34:12 +01:00
|
|
|
}
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
|
2016-06-04 02:02:07 +02:00
|
|
|
RateLimiter* rate_limiter = options_.backup_rate_limiter.get();
|
|
|
|
if (rate_limiter) {
|
2018-09-06 03:07:53 +02:00
|
|
|
copy_file_buffer_size_ = static_cast<size_t>(rate_limiter->GetSingleBurstBytes());
|
2014-03-24 19:38:44 +01:00
|
|
|
}
|
|
|
|
|
2015-07-02 20:35:51 +02:00
|
|
|
// A set into which we will insert the dst_paths that are calculated for live
|
|
|
|
// files and live WAL files.
|
|
|
|
// This is used to check whether a live files shares a dst_path with another
|
|
|
|
// live file.
|
|
|
|
std::unordered_set<std::string> live_dst_paths;
|
2016-03-02 04:33:33 +01:00
|
|
|
|
2016-02-29 21:56:55 +01:00
|
|
|
std::vector<BackupAfterCopyOrCreateWorkItem> backup_items_to_finish;
|
|
|
|
// Add a CopyOrCreateWorkItem to the channel for each live file
|
2017-04-24 23:57:27 +02:00
|
|
|
db->DisableFileDeletions();
|
|
|
|
if (s.ok()) {
|
|
|
|
CheckpointImpl checkpoint(db);
|
|
|
|
uint64_t sequence_number = 0;
|
2018-11-13 20:14:41 +01:00
|
|
|
DBOptions db_options = db->GetDBOptions();
|
|
|
|
EnvOptions src_raw_env_options(db_options);
|
2017-04-24 23:57:27 +02:00
|
|
|
s = checkpoint.CreateCustomCheckpoint(
|
2018-11-13 20:14:41 +01:00
|
|
|
db_options,
|
2018-04-13 02:55:14 +02:00
|
|
|
[&](const std::string& /*src_dirname*/, const std::string& /*fname*/,
|
2017-04-24 23:57:27 +02:00
|
|
|
FileType) {
|
|
|
|
// custom checkpoint will switch to calling copy_file_cb after it sees
|
|
|
|
// NotSupported returned from link_file_cb.
|
|
|
|
return Status::NotSupported();
|
|
|
|
} /* link_file_cb */,
|
|
|
|
[&](const std::string& src_dirname, const std::string& fname,
|
2020-07-03 03:13:31 +02:00
|
|
|
uint64_t size_limit_bytes, FileType type,
|
|
|
|
const std::string& checksum_func_name,
|
|
|
|
const std::string& checksum_val) {
|
2017-04-24 23:57:27 +02:00
|
|
|
if (type == kLogFile && !options_.backup_log_files) {
|
|
|
|
return Status::OK();
|
|
|
|
}
|
|
|
|
Log(options_.info_log, "add file for backup %s", fname.c_str());
|
|
|
|
uint64_t size_bytes = 0;
|
|
|
|
Status st;
|
|
|
|
if (type == kTableFile) {
|
|
|
|
st = db_env_->GetFileSize(src_dirname + fname, &size_bytes);
|
|
|
|
}
|
2018-11-13 20:14:41 +01:00
|
|
|
EnvOptions src_env_options;
|
|
|
|
switch (type) {
|
|
|
|
case kLogFile:
|
|
|
|
src_env_options =
|
|
|
|
db_env_->OptimizeForLogRead(src_raw_env_options);
|
|
|
|
break;
|
|
|
|
case kTableFile:
|
|
|
|
src_env_options = db_env_->OptimizeForCompactionTableRead(
|
|
|
|
src_raw_env_options, ImmutableDBOptions(db_options));
|
|
|
|
break;
|
|
|
|
case kDescriptorFile:
|
|
|
|
src_env_options =
|
|
|
|
db_env_->OptimizeForManifestRead(src_raw_env_options);
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
// Other backed up files (like options file) are not read by live
|
|
|
|
// DB, so don't need to worry about avoiding mixing buffered and
|
|
|
|
// direct I/O. Just use plain defaults.
|
|
|
|
src_env_options = src_raw_env_options;
|
|
|
|
break;
|
|
|
|
}
|
2017-04-24 23:57:27 +02:00
|
|
|
if (st.ok()) {
|
|
|
|
st = AddBackupFileWorkItem(
|
|
|
|
live_dst_paths, backup_items_to_finish, new_backup_id,
|
|
|
|
options_.share_table_files && type == kTableFile, src_dirname,
|
2018-11-13 20:14:41 +01:00
|
|
|
fname, src_env_options, rate_limiter, size_bytes,
|
|
|
|
size_limit_bytes,
|
2017-04-24 23:57:27 +02:00
|
|
|
options_.share_files_with_checksum && type == kTableFile,
|
2020-07-03 03:13:31 +02:00
|
|
|
options.progress_callback, "" /* contents */,
|
|
|
|
checksum_func_name, checksum_val);
|
2017-04-24 23:57:27 +02:00
|
|
|
}
|
|
|
|
return st;
|
|
|
|
} /* copy_file_cb */,
|
|
|
|
[&](const std::string& fname, const std::string& contents, FileType) {
|
|
|
|
Log(options_.info_log, "add file for backup %s", fname.c_str());
|
|
|
|
return AddBackupFileWorkItem(
|
|
|
|
live_dst_paths, backup_items_to_finish, new_backup_id,
|
2018-11-13 20:14:41 +01:00
|
|
|
false /* shared */, "" /* src_dir */, fname,
|
|
|
|
EnvOptions() /* src_env_options */, rate_limiter, contents.size(),
|
|
|
|
0 /* size_limit */, false /* shared_checksum */,
|
2020-03-29 04:05:54 +02:00
|
|
|
options.progress_callback, contents);
|
2017-04-24 23:57:27 +02:00
|
|
|
} /* create_file_cb */,
|
2020-07-03 03:13:31 +02:00
|
|
|
&sequence_number, options.flush_before_backup ? 0 : port::kMaxUint64,
|
2020-08-12 22:28:43 +02:00
|
|
|
db_options.file_checksum_gen_factory == nullptr ? false : true);
|
2017-04-24 23:57:27 +02:00
|
|
|
if (s.ok()) {
|
|
|
|
new_backup->SetSequenceNumber(sequence_number);
|
2015-07-02 20:35:51 +02:00
|
|
|
}
|
|
|
|
}
|
2017-03-16 03:22:52 +01:00
|
|
|
ROCKS_LOG_INFO(options_.info_log, "add files for backup done, wait finish.");
|
2015-07-02 20:35:51 +02:00
|
|
|
Status item_status;
|
|
|
|
for (auto& item : backup_items_to_finish) {
|
|
|
|
item.result.wait();
|
|
|
|
auto result = item.result.get();
|
|
|
|
item_status = result.status;
|
|
|
|
if (item_status.ok() && item.shared && item.needed_to_copy) {
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
item_status =
|
|
|
|
item.backup_env->RenameFile(item.dst_path_tmp, item.dst_path);
|
2015-07-02 20:35:51 +02:00
|
|
|
}
|
|
|
|
if (item_status.ok()) {
|
2020-06-25 04:30:15 +02:00
|
|
|
item_status = new_backup.get()->AddFile(std::make_shared<FileInfo>(
|
2020-08-12 22:28:43 +02:00
|
|
|
item.dst_relative, result.size, result.checksum_hex,
|
|
|
|
result.custom_checksum_hex, result.checksum_func_name, result.db_id,
|
2020-06-25 04:30:15 +02:00
|
|
|
result.db_session_id));
|
2015-07-02 20:35:51 +02:00
|
|
|
}
|
|
|
|
if (!item_status.ok()) {
|
|
|
|
s = item_status;
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
// we copied all the files, enable file deletions
|
2014-05-20 23:28:51 +02:00
|
|
|
db->EnableFileDeletions(false);
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
|
2014-09-09 22:44:42 +02:00
|
|
|
auto backup_time = backup_env_->NowMicros() - start_backup;
|
|
|
|
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
if (s.ok()) {
|
|
|
|
// persist the backup metadata on the disk
|
2015-01-05 14:35:09 +01:00
|
|
|
s = new_backup->StoreToFile(options_.sync);
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
}
|
2014-03-05 02:02:25 +01:00
|
|
|
if (s.ok() && options_.sync) {
|
2018-11-09 20:17:34 +01:00
|
|
|
std::unique_ptr<Directory> backup_private_directory;
|
2014-03-05 02:02:25 +01:00
|
|
|
backup_env_->NewDirectory(
|
|
|
|
GetAbsolutePath(GetPrivateFileRel(new_backup_id, false)),
|
|
|
|
&backup_private_directory);
|
|
|
|
if (backup_private_directory != nullptr) {
|
2018-09-14 22:17:12 +02:00
|
|
|
s = backup_private_directory->Fsync();
|
2014-03-05 02:02:25 +01:00
|
|
|
}
|
2018-09-14 22:17:12 +02:00
|
|
|
if (s.ok() && private_directory_ != nullptr) {
|
|
|
|
s = private_directory_->Fsync();
|
2014-03-05 02:02:25 +01:00
|
|
|
}
|
2018-09-14 22:17:12 +02:00
|
|
|
if (s.ok() && meta_directory_ != nullptr) {
|
|
|
|
s = meta_directory_->Fsync();
|
2014-03-05 02:02:25 +01:00
|
|
|
}
|
2018-09-14 22:17:12 +02:00
|
|
|
if (s.ok() && shared_directory_ != nullptr) {
|
|
|
|
s = shared_directory_->Fsync();
|
2014-03-05 02:02:25 +01:00
|
|
|
}
|
2018-09-14 22:17:12 +02:00
|
|
|
if (s.ok() && backup_directory_ != nullptr) {
|
|
|
|
s = backup_directory_->Fsync();
|
2014-03-05 02:02:25 +01:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2014-09-09 22:44:42 +02:00
|
|
|
if (s.ok()) {
|
|
|
|
backup_statistics_.IncrementNumberSuccessBackup();
|
|
|
|
}
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
if (!s.ok()) {
|
2014-09-09 22:44:42 +02:00
|
|
|
backup_statistics_.IncrementNumberFailBackup();
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
// clean all the files we might have created
|
2017-03-16 03:22:52 +01:00
|
|
|
ROCKS_LOG_INFO(options_.info_log, "Backup failed -- %s",
|
|
|
|
s.ToString().c_str());
|
|
|
|
ROCKS_LOG_INFO(options_.info_log, "Backup Statistics %s\n",
|
|
|
|
backup_statistics_.ToString().c_str());
|
Fix BackupEngine
Summary:
In D28521 we removed GarbageCollect() from BackupEngine's constructor. The reason was that opening BackupEngine on HDFS was very slow and in most cases we didn't have any garbage. We allowed the user to call GarbageCollect() when it detects some garbage files in his backup directory.
Unfortunately, this left us vulnerable to an interesting issue. Let's say we started a backup and copied files {1, 3} but the backup failed. On another host, we restore DB from backup and generate {1, 3, 5}. Since {1, 3} is already there, we will not overwrite. However, these files might be from a different database so their contents might be different. See internal task t6781803 for more info.
Now, when we're copying files and we discover a file already there, we check:
1. if the file is not referenced from any backups, we overwrite the file.
2. if the file is referenced from other backups AND the checksums don't match, we fail the backup. This will only happen if user is using a single backup directory for backing up two different databases.
3. if the file is referenced from other backups AND the checksums match, it's all good. We skip the copy and go copy the next file.
Test Plan: Added new test to backupable_db_test. The test fails before this patch.
Reviewers: sdong, rven, yhchiang
Reviewed By: yhchiang
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D37599
2015-05-08 02:39:19 +02:00
|
|
|
// delete files that we might have already written
|
Auto-GarbageCollect on PurgeOldBackups and DeleteBackup (#6015)
Summary:
Only if there is a crash, power failure, or I/O error in
DeleteBackup, shared or private files from the backup might be left
behind that are not cleaned up by PurgeOldBackups or DeleteBackup-- only
by GarbageCollect. This makes the BackupEngine API "leaky by default."
Even if it means a modest performance hit, I think we should make
Delete and Purge do as they say, with ongoing best effort: i.e. future
calls will attempt to finish any incomplete work from earlier calls.
This change does that by having DeleteBackup and PurgeOldBackups do a
GarbageCollect, unless (to minimize performance hit) this BackupEngine
has already done a GarbageCollect and there have been no
deletion-related I/O errors in that GarbageCollect or since then.
Rejected alternative 1: remove meta file last instead of first. This would in theory turn partially deleted backups into corrupted backups, but code changes would be needed to allow the missing files and consider it acceptably corrupt, rather than failing to open the BackupEngine. This might be a reasonable choice, but I mostly rejected it because it doesn't solve the legacy problem of cleaning up existing lingering files.
Rejected alternative 2: use a deletion marker file. If deletion started with creating a file that marks a backup as flagged for deletion, then we could reliably detect partially deleted backups and efficiently finish removing them. In addition to not solving the legacy problem, this could be precarious if there's a disk full situation, and we try to create a new file in order to delete some files. Ugh.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6015
Test Plan: Updated unit tests
Differential Revision: D18401333
Pulled By: pdillinger
fbshipit-source-id: 12944e372ce6809f3f5a4c416c3b321a8927d925
2019-11-09 04:13:41 +01:00
|
|
|
might_need_garbage_collect_ = true;
|
Fix BackupEngine
Summary:
In D28521 we removed GarbageCollect() from BackupEngine's constructor. The reason was that opening BackupEngine on HDFS was very slow and in most cases we didn't have any garbage. We allowed the user to call GarbageCollect() when it detects some garbage files in his backup directory.
Unfortunately, this left us vulnerable to an interesting issue. Let's say we started a backup and copied files {1, 3} but the backup failed. On another host, we restore DB from backup and generate {1, 3, 5}. Since {1, 3} is already there, we will not overwrite. However, these files might be from a different database so their contents might be different. See internal task t6781803 for more info.
Now, when we're copying files and we discover a file already there, we check:
1. if the file is not referenced from any backups, we overwrite the file.
2. if the file is referenced from other backups AND the checksums don't match, we fail the backup. This will only happen if user is using a single backup directory for backing up two different databases.
3. if the file is referenced from other backups AND the checksums match, it's all good. We skip the copy and go copy the next file.
Test Plan: Added new test to backupable_db_test. The test fails before this patch.
Reviewers: sdong, rven, yhchiang
Reviewed By: yhchiang
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D37599
2015-05-08 02:39:19 +02:00
|
|
|
DeleteBackup(new_backup_id);
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
|
|
|
// here we know that we succeeded and installed the new backup
|
|
|
|
// in the LATEST_BACKUP file
|
|
|
|
latest_backup_id_ = new_backup_id;
|
2017-09-01 00:30:35 +02:00
|
|
|
latest_valid_backup_id_ = new_backup_id;
|
2017-03-16 03:22:52 +01:00
|
|
|
ROCKS_LOG_INFO(options_.info_log, "Backup DONE. All is good");
|
2014-09-09 22:44:42 +02:00
|
|
|
|
|
|
|
// backup_speed is in byte/second
|
2015-01-05 14:35:09 +01:00
|
|
|
double backup_speed = new_backup->GetSize() / (1.048576 * backup_time);
|
2017-03-16 03:22:52 +01:00
|
|
|
ROCKS_LOG_INFO(options_.info_log, "Backup number of files: %u",
|
|
|
|
new_backup->GetNumberFiles());
|
2015-02-27 23:03:56 +01:00
|
|
|
char human_size[16];
|
|
|
|
AppendHumanBytes(new_backup->GetSize(), human_size, sizeof(human_size));
|
2017-03-16 03:22:52 +01:00
|
|
|
ROCKS_LOG_INFO(options_.info_log, "Backup size: %s", human_size);
|
|
|
|
ROCKS_LOG_INFO(options_.info_log, "Backup time: %" PRIu64 " microseconds",
|
|
|
|
backup_time);
|
|
|
|
ROCKS_LOG_INFO(options_.info_log, "Backup speed: %.3f MB/s", backup_speed);
|
|
|
|
ROCKS_LOG_INFO(options_.info_log, "Backup Statistics %s",
|
|
|
|
backup_statistics_.ToString().c_str());
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
2014-01-28 20:26:07 +01:00
|
|
|
Status BackupEngineImpl::PurgeOldBackups(uint32_t num_backups_to_keep) {
|
2015-07-14 20:51:36 +02:00
|
|
|
assert(initialized_);
|
2014-04-25 21:49:29 +02:00
|
|
|
assert(!read_only_);
|
Auto-GarbageCollect on PurgeOldBackups and DeleteBackup (#6015)
Summary:
Only if there is a crash, power failure, or I/O error in
DeleteBackup, shared or private files from the backup might be left
behind that are not cleaned up by PurgeOldBackups or DeleteBackup-- only
by GarbageCollect. This makes the BackupEngine API "leaky by default."
Even if it means a modest performance hit, I think we should make
Delete and Purge do as they say, with ongoing best effort: i.e. future
calls will attempt to finish any incomplete work from earlier calls.
This change does that by having DeleteBackup and PurgeOldBackups do a
GarbageCollect, unless (to minimize performance hit) this BackupEngine
has already done a GarbageCollect and there have been no
deletion-related I/O errors in that GarbageCollect or since then.
Rejected alternative 1: remove meta file last instead of first. This would in theory turn partially deleted backups into corrupted backups, but code changes would be needed to allow the missing files and consider it acceptably corrupt, rather than failing to open the BackupEngine. This might be a reasonable choice, but I mostly rejected it because it doesn't solve the legacy problem of cleaning up existing lingering files.
Rejected alternative 2: use a deletion marker file. If deletion started with creating a file that marks a backup as flagged for deletion, then we could reliably detect partially deleted backups and efficiently finish removing them. In addition to not solving the legacy problem, this could be precarious if there's a disk full situation, and we try to create a new file in order to delete some files. Ugh.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6015
Test Plan: Updated unit tests
Differential Revision: D18401333
Pulled By: pdillinger
fbshipit-source-id: 12944e372ce6809f3f5a4c416c3b321a8927d925
2019-11-09 04:13:41 +01:00
|
|
|
|
|
|
|
// Best effort deletion even with errors
|
|
|
|
Status overall_status = Status::OK();
|
|
|
|
|
2017-03-16 03:22:52 +01:00
|
|
|
ROCKS_LOG_INFO(options_.info_log, "Purging old backups, keeping %u",
|
|
|
|
num_backups_to_keep);
|
2014-11-13 18:51:41 +01:00
|
|
|
std::vector<BackupID> to_delete;
|
|
|
|
auto itr = backups_.begin();
|
|
|
|
while ((backups_.size() - to_delete.size()) > num_backups_to_keep) {
|
|
|
|
to_delete.push_back(itr->first);
|
|
|
|
itr++;
|
|
|
|
}
|
|
|
|
for (auto backup_id : to_delete) {
|
Auto-GarbageCollect on PurgeOldBackups and DeleteBackup (#6015)
Summary:
Only if there is a crash, power failure, or I/O error in
DeleteBackup, shared or private files from the backup might be left
behind that are not cleaned up by PurgeOldBackups or DeleteBackup-- only
by GarbageCollect. This makes the BackupEngine API "leaky by default."
Even if it means a modest performance hit, I think we should make
Delete and Purge do as they say, with ongoing best effort: i.e. future
calls will attempt to finish any incomplete work from earlier calls.
This change does that by having DeleteBackup and PurgeOldBackups do a
GarbageCollect, unless (to minimize performance hit) this BackupEngine
has already done a GarbageCollect and there have been no
deletion-related I/O errors in that GarbageCollect or since then.
Rejected alternative 1: remove meta file last instead of first. This would in theory turn partially deleted backups into corrupted backups, but code changes would be needed to allow the missing files and consider it acceptably corrupt, rather than failing to open the BackupEngine. This might be a reasonable choice, but I mostly rejected it because it doesn't solve the legacy problem of cleaning up existing lingering files.
Rejected alternative 2: use a deletion marker file. If deletion started with creating a file that marks a backup as flagged for deletion, then we could reliably detect partially deleted backups and efficiently finish removing them. In addition to not solving the legacy problem, this could be precarious if there's a disk full situation, and we try to create a new file in order to delete some files. Ugh.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6015
Test Plan: Updated unit tests
Differential Revision: D18401333
Pulled By: pdillinger
fbshipit-source-id: 12944e372ce6809f3f5a4c416c3b321a8927d925
2019-11-09 04:13:41 +01:00
|
|
|
auto s = DeleteBackupInternal(backup_id);
|
2015-07-14 20:51:36 +02:00
|
|
|
if (!s.ok()) {
|
Auto-GarbageCollect on PurgeOldBackups and DeleteBackup (#6015)
Summary:
Only if there is a crash, power failure, or I/O error in
DeleteBackup, shared or private files from the backup might be left
behind that are not cleaned up by PurgeOldBackups or DeleteBackup-- only
by GarbageCollect. This makes the BackupEngine API "leaky by default."
Even if it means a modest performance hit, I think we should make
Delete and Purge do as they say, with ongoing best effort: i.e. future
calls will attempt to finish any incomplete work from earlier calls.
This change does that by having DeleteBackup and PurgeOldBackups do a
GarbageCollect, unless (to minimize performance hit) this BackupEngine
has already done a GarbageCollect and there have been no
deletion-related I/O errors in that GarbageCollect or since then.
Rejected alternative 1: remove meta file last instead of first. This would in theory turn partially deleted backups into corrupted backups, but code changes would be needed to allow the missing files and consider it acceptably corrupt, rather than failing to open the BackupEngine. This might be a reasonable choice, but I mostly rejected it because it doesn't solve the legacy problem of cleaning up existing lingering files.
Rejected alternative 2: use a deletion marker file. If deletion started with creating a file that marks a backup as flagged for deletion, then we could reliably detect partially deleted backups and efficiently finish removing them. In addition to not solving the legacy problem, this could be precarious if there's a disk full situation, and we try to create a new file in order to delete some files. Ugh.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6015
Test Plan: Updated unit tests
Differential Revision: D18401333
Pulled By: pdillinger
fbshipit-source-id: 12944e372ce6809f3f5a4c416c3b321a8927d925
2019-11-09 04:13:41 +01:00
|
|
|
overall_status = s;
|
2015-07-14 20:51:36 +02:00
|
|
|
}
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
}
|
Auto-GarbageCollect on PurgeOldBackups and DeleteBackup (#6015)
Summary:
Only if there is a crash, power failure, or I/O error in
DeleteBackup, shared or private files from the backup might be left
behind that are not cleaned up by PurgeOldBackups or DeleteBackup-- only
by GarbageCollect. This makes the BackupEngine API "leaky by default."
Even if it means a modest performance hit, I think we should make
Delete and Purge do as they say, with ongoing best effort: i.e. future
calls will attempt to finish any incomplete work from earlier calls.
This change does that by having DeleteBackup and PurgeOldBackups do a
GarbageCollect, unless (to minimize performance hit) this BackupEngine
has already done a GarbageCollect and there have been no
deletion-related I/O errors in that GarbageCollect or since then.
Rejected alternative 1: remove meta file last instead of first. This would in theory turn partially deleted backups into corrupted backups, but code changes would be needed to allow the missing files and consider it acceptably corrupt, rather than failing to open the BackupEngine. This might be a reasonable choice, but I mostly rejected it because it doesn't solve the legacy problem of cleaning up existing lingering files.
Rejected alternative 2: use a deletion marker file. If deletion started with creating a file that marks a backup as flagged for deletion, then we could reliably detect partially deleted backups and efficiently finish removing them. In addition to not solving the legacy problem, this could be precarious if there's a disk full situation, and we try to create a new file in order to delete some files. Ugh.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6015
Test Plan: Updated unit tests
Differential Revision: D18401333
Pulled By: pdillinger
fbshipit-source-id: 12944e372ce6809f3f5a4c416c3b321a8927d925
2019-11-09 04:13:41 +01:00
|
|
|
// Clean up after any incomplete backup deletion, potentially from
|
|
|
|
// earlier session.
|
|
|
|
if (might_need_garbage_collect_) {
|
|
|
|
auto s = GarbageCollect();
|
|
|
|
if (!s.ok() && overall_status.ok()) {
|
|
|
|
overall_status = s;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return overall_status;
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
}
|
|
|
|
|
2014-01-28 20:26:07 +01:00
|
|
|
Status BackupEngineImpl::DeleteBackup(BackupID backup_id) {
|
Auto-GarbageCollect on PurgeOldBackups and DeleteBackup (#6015)
Summary:
Only if there is a crash, power failure, or I/O error in
DeleteBackup, shared or private files from the backup might be left
behind that are not cleaned up by PurgeOldBackups or DeleteBackup-- only
by GarbageCollect. This makes the BackupEngine API "leaky by default."
Even if it means a modest performance hit, I think we should make
Delete and Purge do as they say, with ongoing best effort: i.e. future
calls will attempt to finish any incomplete work from earlier calls.
This change does that by having DeleteBackup and PurgeOldBackups do a
GarbageCollect, unless (to minimize performance hit) this BackupEngine
has already done a GarbageCollect and there have been no
deletion-related I/O errors in that GarbageCollect or since then.
Rejected alternative 1: remove meta file last instead of first. This would in theory turn partially deleted backups into corrupted backups, but code changes would be needed to allow the missing files and consider it acceptably corrupt, rather than failing to open the BackupEngine. This might be a reasonable choice, but I mostly rejected it because it doesn't solve the legacy problem of cleaning up existing lingering files.
Rejected alternative 2: use a deletion marker file. If deletion started with creating a file that marks a backup as flagged for deletion, then we could reliably detect partially deleted backups and efficiently finish removing them. In addition to not solving the legacy problem, this could be precarious if there's a disk full situation, and we try to create a new file in order to delete some files. Ugh.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6015
Test Plan: Updated unit tests
Differential Revision: D18401333
Pulled By: pdillinger
fbshipit-source-id: 12944e372ce6809f3f5a4c416c3b321a8927d925
2019-11-09 04:13:41 +01:00
|
|
|
auto s1 = DeleteBackupInternal(backup_id);
|
|
|
|
auto s2 = Status::OK();
|
|
|
|
|
|
|
|
// Clean up after any incomplete backup deletion, potentially from
|
|
|
|
// earlier session.
|
|
|
|
if (might_need_garbage_collect_) {
|
|
|
|
s2 = GarbageCollect();
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!s1.ok()) {
|
|
|
|
return s1;
|
|
|
|
} else {
|
|
|
|
return s2;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
// Does not auto-GarbageCollect
|
|
|
|
Status BackupEngineImpl::DeleteBackupInternal(BackupID backup_id) {
|
2015-07-14 20:51:36 +02:00
|
|
|
assert(initialized_);
|
2014-04-25 21:49:29 +02:00
|
|
|
assert(!read_only_);
|
Auto-GarbageCollect on PurgeOldBackups and DeleteBackup (#6015)
Summary:
Only if there is a crash, power failure, or I/O error in
DeleteBackup, shared or private files from the backup might be left
behind that are not cleaned up by PurgeOldBackups or DeleteBackup-- only
by GarbageCollect. This makes the BackupEngine API "leaky by default."
Even if it means a modest performance hit, I think we should make
Delete and Purge do as they say, with ongoing best effort: i.e. future
calls will attempt to finish any incomplete work from earlier calls.
This change does that by having DeleteBackup and PurgeOldBackups do a
GarbageCollect, unless (to minimize performance hit) this BackupEngine
has already done a GarbageCollect and there have been no
deletion-related I/O errors in that GarbageCollect or since then.
Rejected alternative 1: remove meta file last instead of first. This would in theory turn partially deleted backups into corrupted backups, but code changes would be needed to allow the missing files and consider it acceptably corrupt, rather than failing to open the BackupEngine. This might be a reasonable choice, but I mostly rejected it because it doesn't solve the legacy problem of cleaning up existing lingering files.
Rejected alternative 2: use a deletion marker file. If deletion started with creating a file that marks a backup as flagged for deletion, then we could reliably detect partially deleted backups and efficiently finish removing them. In addition to not solving the legacy problem, this could be precarious if there's a disk full situation, and we try to create a new file in order to delete some files. Ugh.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6015
Test Plan: Updated unit tests
Differential Revision: D18401333
Pulled By: pdillinger
fbshipit-source-id: 12944e372ce6809f3f5a4c416c3b321a8927d925
2019-11-09 04:13:41 +01:00
|
|
|
|
2017-03-16 03:22:52 +01:00
|
|
|
ROCKS_LOG_INFO(options_.info_log, "Deleting backup %u", backup_id);
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
auto backup = backups_.find(backup_id);
|
2014-11-13 18:51:41 +01:00
|
|
|
if (backup != backups_.end()) {
|
2015-07-14 20:51:36 +02:00
|
|
|
auto s = backup->second->Delete();
|
|
|
|
if (!s.ok()) {
|
|
|
|
return s;
|
|
|
|
}
|
2014-11-13 18:51:41 +01:00
|
|
|
backups_.erase(backup);
|
|
|
|
} else {
|
|
|
|
auto corrupt = corrupt_backups_.find(backup_id);
|
|
|
|
if (corrupt == corrupt_backups_.end()) {
|
|
|
|
return Status::NotFound("Backup not found");
|
|
|
|
}
|
2015-07-14 20:51:36 +02:00
|
|
|
auto s = corrupt->second.second->Delete();
|
|
|
|
if (!s.ok()) {
|
|
|
|
return s;
|
|
|
|
}
|
2014-11-13 18:51:41 +01:00
|
|
|
corrupt_backups_.erase(corrupt);
|
|
|
|
}
|
|
|
|
|
Auto-GarbageCollect on PurgeOldBackups and DeleteBackup (#6015)
Summary:
Only if there is a crash, power failure, or I/O error in
DeleteBackup, shared or private files from the backup might be left
behind that are not cleaned up by PurgeOldBackups or DeleteBackup-- only
by GarbageCollect. This makes the BackupEngine API "leaky by default."
Even if it means a modest performance hit, I think we should make
Delete and Purge do as they say, with ongoing best effort: i.e. future
calls will attempt to finish any incomplete work from earlier calls.
This change does that by having DeleteBackup and PurgeOldBackups do a
GarbageCollect, unless (to minimize performance hit) this BackupEngine
has already done a GarbageCollect and there have been no
deletion-related I/O errors in that GarbageCollect or since then.
Rejected alternative 1: remove meta file last instead of first. This would in theory turn partially deleted backups into corrupted backups, but code changes would be needed to allow the missing files and consider it acceptably corrupt, rather than failing to open the BackupEngine. This might be a reasonable choice, but I mostly rejected it because it doesn't solve the legacy problem of cleaning up existing lingering files.
Rejected alternative 2: use a deletion marker file. If deletion started with creating a file that marks a backup as flagged for deletion, then we could reliably detect partially deleted backups and efficiently finish removing them. In addition to not solving the legacy problem, this could be precarious if there's a disk full situation, and we try to create a new file in order to delete some files. Ugh.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6015
Test Plan: Updated unit tests
Differential Revision: D18401333
Pulled By: pdillinger
fbshipit-source-id: 12944e372ce6809f3f5a4c416c3b321a8927d925
2019-11-09 04:13:41 +01:00
|
|
|
// After removing meta file, best effort deletion even with errors.
|
|
|
|
// (Don't delete other files if we can't delete the meta file right
|
|
|
|
// now.)
|
2019-11-25 23:18:10 +01:00
|
|
|
std::vector<std::string> to_delete;
|
|
|
|
for (auto& itr : backuped_file_infos_) {
|
|
|
|
if (itr.second->refs == 0) {
|
|
|
|
Status s = backup_env_->DeleteFile(GetAbsolutePath(itr.first));
|
|
|
|
ROCKS_LOG_INFO(options_.info_log, "Deleting %s -- %s", itr.first.c_str(),
|
|
|
|
s.ToString().c_str());
|
|
|
|
to_delete.push_back(itr.first);
|
|
|
|
if (!s.ok()) {
|
|
|
|
// Trying again later might work
|
|
|
|
might_need_garbage_collect_ = true;
|
2018-04-06 05:58:35 +02:00
|
|
|
}
|
2014-11-13 18:51:41 +01:00
|
|
|
}
|
2019-11-25 23:18:10 +01:00
|
|
|
}
|
|
|
|
for (auto& td : to_delete) {
|
|
|
|
backuped_file_infos_.erase(td);
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
}
|
2014-11-13 18:51:41 +01:00
|
|
|
|
|
|
|
// take care of private dirs -- GarbageCollect() will take care of them
|
|
|
|
// if they are not empty
|
|
|
|
std::string private_dir = GetPrivateFileRel(backup_id);
|
|
|
|
Status s = backup_env_->DeleteDir(GetAbsolutePath(private_dir));
|
2017-03-16 03:22:52 +01:00
|
|
|
ROCKS_LOG_INFO(options_.info_log, "Deleting private dir %s -- %s",
|
|
|
|
private_dir.c_str(), s.ToString().c_str());
|
Auto-GarbageCollect on PurgeOldBackups and DeleteBackup (#6015)
Summary:
Only if there is a crash, power failure, or I/O error in
DeleteBackup, shared or private files from the backup might be left
behind that are not cleaned up by PurgeOldBackups or DeleteBackup-- only
by GarbageCollect. This makes the BackupEngine API "leaky by default."
Even if it means a modest performance hit, I think we should make
Delete and Purge do as they say, with ongoing best effort: i.e. future
calls will attempt to finish any incomplete work from earlier calls.
This change does that by having DeleteBackup and PurgeOldBackups do a
GarbageCollect, unless (to minimize performance hit) this BackupEngine
has already done a GarbageCollect and there have been no
deletion-related I/O errors in that GarbageCollect or since then.
Rejected alternative 1: remove meta file last instead of first. This would in theory turn partially deleted backups into corrupted backups, but code changes would be needed to allow the missing files and consider it acceptably corrupt, rather than failing to open the BackupEngine. This might be a reasonable choice, but I mostly rejected it because it doesn't solve the legacy problem of cleaning up existing lingering files.
Rejected alternative 2: use a deletion marker file. If deletion started with creating a file that marks a backup as flagged for deletion, then we could reliably detect partially deleted backups and efficiently finish removing them. In addition to not solving the legacy problem, this could be precarious if there's a disk full situation, and we try to create a new file in order to delete some files. Ugh.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6015
Test Plan: Updated unit tests
Differential Revision: D18401333
Pulled By: pdillinger
fbshipit-source-id: 12944e372ce6809f3f5a4c416c3b321a8927d925
2019-11-09 04:13:41 +01:00
|
|
|
if (!s.ok()) {
|
|
|
|
// Full gc or trying again later might work
|
|
|
|
might_need_garbage_collect_ = true;
|
|
|
|
}
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
return Status::OK();
|
|
|
|
}
|
|
|
|
|
2014-01-28 20:26:07 +01:00
|
|
|
void BackupEngineImpl::GetBackupInfo(std::vector<BackupInfo>* backup_info) {
|
2015-07-14 20:51:36 +02:00
|
|
|
assert(initialized_);
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
backup_info->reserve(backups_.size());
|
|
|
|
for (auto& backup : backups_) {
|
2015-01-05 14:35:09 +01:00
|
|
|
if (!backup.second->Empty()) {
|
2016-04-01 19:56:52 +02:00
|
|
|
backup_info->push_back(BackupInfo(
|
|
|
|
backup.first, backup.second->GetTimestamp(), backup.second->GetSize(),
|
|
|
|
backup.second->GetNumberFiles(), backup.second->GetAppMetadata()));
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2014-11-13 18:51:41 +01:00
|
|
|
void
|
|
|
|
BackupEngineImpl::GetCorruptedBackups(
|
|
|
|
std::vector<BackupID>* corrupt_backup_ids) {
|
2015-07-14 20:51:36 +02:00
|
|
|
assert(initialized_);
|
2014-11-13 18:51:41 +01:00
|
|
|
corrupt_backup_ids->reserve(corrupt_backups_.size());
|
|
|
|
for (auto& backup : corrupt_backups_) {
|
|
|
|
corrupt_backup_ids->push_back(backup.first);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2020-03-29 04:05:54 +02:00
|
|
|
Status BackupEngineImpl::RestoreDBFromBackup(const RestoreOptions& options,
|
|
|
|
BackupID backup_id,
|
|
|
|
const std::string& db_dir,
|
|
|
|
const std::string& wal_dir) {
|
2015-07-14 20:51:36 +02:00
|
|
|
assert(initialized_);
|
2014-11-13 18:51:41 +01:00
|
|
|
auto corrupt_itr = corrupt_backups_.find(backup_id);
|
|
|
|
if (corrupt_itr != corrupt_backups_.end()) {
|
|
|
|
return corrupt_itr->second.first;
|
|
|
|
}
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
auto backup_itr = backups_.find(backup_id);
|
|
|
|
if (backup_itr == backups_.end()) {
|
|
|
|
return Status::NotFound("Backup not found");
|
|
|
|
}
|
|
|
|
auto& backup = backup_itr->second;
|
2015-01-05 14:35:09 +01:00
|
|
|
if (backup->Empty()) {
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
return Status::NotFound("Backup not found");
|
|
|
|
}
|
|
|
|
|
2017-03-16 03:22:52 +01:00
|
|
|
ROCKS_LOG_INFO(options_.info_log, "Restoring backup id %u\n", backup_id);
|
|
|
|
ROCKS_LOG_INFO(options_.info_log, "keep_log_files: %d\n",
|
2020-03-29 04:05:54 +02:00
|
|
|
static_cast<int>(options.keep_log_files));
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
|
|
|
|
// just in case. Ignore errors
|
|
|
|
db_env_->CreateDirIfMissing(db_dir);
|
|
|
|
db_env_->CreateDirIfMissing(wal_dir);
|
|
|
|
|
2020-03-29 04:05:54 +02:00
|
|
|
if (options.keep_log_files) {
|
2014-03-17 23:39:23 +01:00
|
|
|
// delete files in db_dir, but keep all the log files
|
|
|
|
DeleteChildren(db_dir, 1 << kLogFile);
|
|
|
|
// move all the files from archive dir to wal_dir
|
|
|
|
std::string archive_dir = ArchivalDirectory(wal_dir);
|
|
|
|
std::vector<std::string> archive_files;
|
|
|
|
db_env_->GetChildren(archive_dir, &archive_files); // ignore errors
|
|
|
|
for (const auto& f : archive_files) {
|
|
|
|
uint64_t number;
|
|
|
|
FileType type;
|
|
|
|
bool ok = ParseFileName(f, &number, &type);
|
|
|
|
if (ok && type == kLogFile) {
|
2017-03-16 03:22:52 +01:00
|
|
|
ROCKS_LOG_INFO(options_.info_log,
|
|
|
|
"Moving log file from archive/ to wal_dir: %s",
|
|
|
|
f.c_str());
|
2014-03-17 23:39:23 +01:00
|
|
|
Status s =
|
|
|
|
db_env_->RenameFile(archive_dir + "/" + f, wal_dir + "/" + f);
|
|
|
|
if (!s.ok()) {
|
|
|
|
// if we can't move log file from archive_dir to wal_dir,
|
|
|
|
// we should fail, since it might mean data loss
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
DeleteChildren(wal_dir);
|
|
|
|
DeleteChildren(ArchivalDirectory(wal_dir));
|
|
|
|
DeleteChildren(db_dir);
|
2013-12-12 23:57:18 +01:00
|
|
|
}
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
|
2020-08-12 22:28:43 +02:00
|
|
|
Status s;
|
|
|
|
// Try to obtain checksum info from backuped DB MANIFEST
|
|
|
|
// The checksum info will be used for validating the checksums of the table
|
|
|
|
// files after restoration, in addtion to the default backup engine crc32c
|
|
|
|
// checksums.
|
|
|
|
std::unique_ptr<FileChecksumList> checksum_list(NewFileChecksumList());
|
|
|
|
s = GetFileChecksumsFromManifestInBackup(backup_env_, backup_id, backup.get(),
|
|
|
|
checksum_list.get());
|
|
|
|
if (!s.ok()) {
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
2016-06-04 02:02:07 +02:00
|
|
|
RateLimiter* rate_limiter = options_.restore_rate_limiter.get();
|
|
|
|
if (rate_limiter) {
|
2020-08-12 22:28:43 +02:00
|
|
|
copy_file_buffer_size_ =
|
|
|
|
static_cast<size_t>(rate_limiter->GetSingleBurstBytes());
|
2014-03-24 19:38:44 +01:00
|
|
|
}
|
2016-02-29 21:56:55 +01:00
|
|
|
std::vector<RestoreAfterCopyOrCreateWorkItem> restore_items_to_finish;
|
2015-01-05 14:35:09 +01:00
|
|
|
for (const auto& file_info : backup->GetFiles()) {
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
const std::string& file = file_info->filename;
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
std::string dst;
|
|
|
|
uint64_t number;
|
|
|
|
FileType type;
|
2020-08-12 22:28:43 +02:00
|
|
|
s = GetFileNameInfo(file, dst, number, type);
|
|
|
|
if (!s.ok()) {
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
|
|
|
std::string src_checksum_func_name = kUnknownFileChecksumFuncName;
|
|
|
|
std::string src_checksum_str = kUnknownFileChecksum;
|
|
|
|
std::string src_checksum_hex;
|
|
|
|
bool has_manifest_checksum = false;
|
|
|
|
if (type == kTableFile) {
|
|
|
|
Status file_checksum_status = checksum_list->SearchOneFileChecksum(
|
|
|
|
number, &src_checksum_str, &src_checksum_func_name);
|
|
|
|
if (file_checksum_status.ok() &&
|
|
|
|
src_checksum_str != kUnknownFileChecksum &&
|
|
|
|
src_checksum_func_name != kUnknownFileChecksumFuncName) {
|
|
|
|
src_checksum_hex = ChecksumStrToHex(src_checksum_str);
|
|
|
|
has_manifest_checksum = true;
|
|
|
|
}
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
}
|
2020-08-12 22:28:43 +02:00
|
|
|
|
|
|
|
// Construct the final path
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
// kLogFile lives in wal_dir and all the rest live in db_dir
|
|
|
|
dst = ((type == kLogFile) ? wal_dir : db_dir) +
|
|
|
|
"/" + dst;
|
|
|
|
|
2017-03-16 03:22:52 +01:00
|
|
|
ROCKS_LOG_INFO(options_.info_log, "Restoring %s to %s\n", file.c_str(),
|
|
|
|
dst.c_str());
|
2020-08-12 22:28:43 +02:00
|
|
|
|
|
|
|
std::string backup_checksum_func_name = file_info->checksum_func_name;
|
|
|
|
std::unique_ptr<FileChecksumGenerator> checksum_func;
|
|
|
|
if (src_checksum_func_name != kUnknownFileChecksumFuncName) {
|
|
|
|
s = SetChecksumGenerator(src_checksum_func_name, checksum_func);
|
|
|
|
if (!s.ok()) {
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
if (checksum_func != nullptr) {
|
|
|
|
backup_checksum_func_name = checksum_func->Name();
|
|
|
|
}
|
|
|
|
}
|
2016-02-29 21:56:55 +01:00
|
|
|
CopyOrCreateWorkItem copy_or_create_work_item(
|
|
|
|
GetAbsolutePath(file), dst, "" /* contents */, backup_env_, db_env_,
|
2018-11-13 20:14:41 +01:00
|
|
|
EnvOptions() /* src_env_options */, false, rate_limiter,
|
2020-08-12 22:28:43 +02:00
|
|
|
0 /* size_limit */, []() {} /* progress_callback */,
|
|
|
|
has_manifest_checksum, src_checksum_func_name, src_checksum_hex,
|
|
|
|
backup_checksum_func_name);
|
2016-02-29 21:56:55 +01:00
|
|
|
RestoreAfterCopyOrCreateWorkItem after_copy_or_create_work_item(
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
copy_or_create_work_item.result.get_future(), file_info->checksum_hex);
|
2016-02-29 21:56:55 +01:00
|
|
|
files_to_copy_or_create_.write(std::move(copy_or_create_work_item));
|
|
|
|
restore_items_to_finish.push_back(
|
|
|
|
std::move(after_copy_or_create_work_item));
|
2015-07-02 20:35:51 +02:00
|
|
|
}
|
|
|
|
Status item_status;
|
|
|
|
for (auto& item : restore_items_to_finish) {
|
|
|
|
item.result.wait();
|
|
|
|
auto result = item.result.get();
|
|
|
|
item_status = result.status;
|
|
|
|
// Note: It is possible that both of the following bad-status cases occur
|
|
|
|
// during copying. But, we only return one status.
|
|
|
|
if (!item_status.ok()) {
|
|
|
|
s = item_status;
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
break;
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
} else if (item.checksum_hex != result.checksum_hex) {
|
2020-08-12 22:28:43 +02:00
|
|
|
// Compare crc32c checksums (especially for non-table files)
|
|
|
|
std::string checksum_info("Expected checksum is " + item.checksum_hex +
|
|
|
|
" while computed checksum is " +
|
|
|
|
result.checksum_hex);
|
|
|
|
s = Status::Corruption("Crc32c checksum check failed: " + checksum_info);
|
2014-01-28 18:43:36 +01:00
|
|
|
break;
|
|
|
|
}
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
}
|
|
|
|
|
2017-03-16 03:22:52 +01:00
|
|
|
ROCKS_LOG_INFO(options_.info_log, "Restoring done -- %s\n",
|
|
|
|
s.ToString().c_str());
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
2020-06-26 20:39:43 +02:00
|
|
|
Status BackupEngineImpl::VerifyBackup(BackupID backup_id,
|
|
|
|
bool verify_with_checksum) {
|
|
|
|
// Check if backup_id is corrupted, or valid and registered
|
2015-09-02 18:03:55 +02:00
|
|
|
assert(initialized_);
|
|
|
|
auto corrupt_itr = corrupt_backups_.find(backup_id);
|
|
|
|
if (corrupt_itr != corrupt_backups_.end()) {
|
|
|
|
return corrupt_itr->second.first;
|
|
|
|
}
|
|
|
|
|
|
|
|
auto backup_itr = backups_.find(backup_id);
|
|
|
|
if (backup_itr == backups_.end()) {
|
|
|
|
return Status::NotFound();
|
|
|
|
}
|
|
|
|
|
|
|
|
auto& backup = backup_itr->second;
|
|
|
|
if (backup->Empty()) {
|
|
|
|
return Status::NotFound();
|
|
|
|
}
|
|
|
|
|
2017-03-16 03:22:52 +01:00
|
|
|
ROCKS_LOG_INFO(options_.info_log, "Verifying backup id %u\n", backup_id);
|
2015-09-02 18:03:55 +02:00
|
|
|
|
2020-06-26 20:39:43 +02:00
|
|
|
// Find all existing backup files belong to backup_id
|
2016-03-02 04:33:33 +01:00
|
|
|
std::unordered_map<std::string, uint64_t> curr_abs_path_to_size;
|
|
|
|
for (const auto& rel_dir : {GetPrivateFileRel(backup_id), GetSharedFileRel(),
|
|
|
|
GetSharedFileWithChecksumRel()}) {
|
|
|
|
const auto abs_dir = GetAbsolutePath(rel_dir);
|
2016-04-26 21:33:30 +02:00
|
|
|
InsertPathnameToSizeBytes(abs_dir, backup_env_, &curr_abs_path_to_size);
|
2016-03-02 04:33:33 +01:00
|
|
|
}
|
|
|
|
|
2020-08-12 22:28:43 +02:00
|
|
|
Status s;
|
|
|
|
std::unique_ptr<FileChecksumList> checksum_list(NewFileChecksumList());
|
|
|
|
if (verify_with_checksum) {
|
|
|
|
// Try to obtain checksum info from backuped DB MANIFEST
|
|
|
|
s = GetFileChecksumsFromManifestInBackup(backup_env_, backup_id,
|
|
|
|
backup.get(), checksum_list.get());
|
|
|
|
if (!s.ok()) {
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2020-06-26 20:39:43 +02:00
|
|
|
// For all files registered in backup
|
2015-09-02 18:03:55 +02:00
|
|
|
for (const auto& file_info : backup->GetFiles()) {
|
2016-03-02 04:33:33 +01:00
|
|
|
const auto abs_path = GetAbsolutePath(file_info->filename);
|
2020-06-26 20:39:43 +02:00
|
|
|
// check existence of the file
|
2016-03-02 04:33:33 +01:00
|
|
|
if (curr_abs_path_to_size.find(abs_path) == curr_abs_path_to_size.end()) {
|
|
|
|
return Status::NotFound("File missing: " + abs_path);
|
2015-09-02 18:03:55 +02:00
|
|
|
}
|
2020-06-26 20:39:43 +02:00
|
|
|
// verify file size
|
2016-03-02 04:33:33 +01:00
|
|
|
if (file_info->size != curr_abs_path_to_size[abs_path]) {
|
2020-06-26 20:39:43 +02:00
|
|
|
std::string size_info("Expected file size is " +
|
|
|
|
ToString(file_info->size) +
|
|
|
|
" while found file size is " +
|
|
|
|
ToString(curr_abs_path_to_size[abs_path]));
|
|
|
|
return Status::Corruption("File corrupted: File size mismatch for " +
|
|
|
|
abs_path + ": " + size_info);
|
|
|
|
}
|
|
|
|
if (verify_with_checksum) {
|
|
|
|
// verify file checksum
|
2020-08-12 22:28:43 +02:00
|
|
|
// try setting checksum_func
|
|
|
|
std::unique_ptr<FileChecksumGenerator> checksum_func;
|
|
|
|
std::string src_checksum_func_name = kUnknownFileChecksumFuncName;
|
|
|
|
std::string src_checksum_str = kUnknownFileChecksum;
|
|
|
|
std::string src_checksum_hex;
|
|
|
|
if (IsSstFile(file_info->filename)) {
|
|
|
|
const std::string& file = file_info->filename;
|
|
|
|
std::string local_name;
|
|
|
|
uint64_t number;
|
|
|
|
FileType type;
|
|
|
|
s = GetFileNameInfo(file, local_name, number, type);
|
|
|
|
if (!s.ok()) {
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
assert(type == kTableFile);
|
|
|
|
|
|
|
|
// Try to get checksum for the table file
|
|
|
|
Status file_checksum_status = checksum_list->SearchOneFileChecksum(
|
|
|
|
number, &src_checksum_str, &src_checksum_func_name);
|
|
|
|
if (file_checksum_status.ok() &&
|
|
|
|
src_checksum_str != kUnknownFileChecksum &&
|
|
|
|
src_checksum_func_name != kUnknownFileChecksumFuncName) {
|
|
|
|
s = SetChecksumGenerator(src_checksum_func_name, checksum_func);
|
|
|
|
if (!s.ok()) {
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
src_checksum_hex = ChecksumStrToHex(src_checksum_str);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2020-06-26 20:39:43 +02:00
|
|
|
ROCKS_LOG_INFO(options_.info_log, "Verifying %s checksum...\n",
|
|
|
|
abs_path.c_str());
|
2020-08-12 22:28:43 +02:00
|
|
|
std::string checksum_hex;
|
|
|
|
std::string custom_checksum_hex;
|
2020-06-26 20:39:43 +02:00
|
|
|
CalculateChecksum(abs_path, backup_env_, EnvOptions(), 0 /* size_limit */,
|
2020-08-12 22:28:43 +02:00
|
|
|
&checksum_hex, checksum_func, &custom_checksum_hex);
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
if (file_info->checksum_hex != checksum_hex) {
|
2020-06-26 20:39:43 +02:00
|
|
|
std::string checksum_info(
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
"Expected checksum is " + file_info->checksum_hex +
|
|
|
|
" while computed checksum is " + checksum_hex);
|
2020-08-12 22:28:43 +02:00
|
|
|
return Status::Corruption("File corrupted: crc32c mismatch for " +
|
2020-06-26 20:39:43 +02:00
|
|
|
abs_path + ": " + checksum_info);
|
|
|
|
}
|
2020-08-12 22:28:43 +02:00
|
|
|
if (checksum_func != nullptr && src_checksum_hex != custom_checksum_hex) {
|
|
|
|
std::string checksum_info("Expected checksum is " + src_checksum_hex +
|
|
|
|
" while computed checksum is " +
|
|
|
|
custom_checksum_hex);
|
|
|
|
return Status::Corruption("File corrupted: " + src_checksum_func_name +
|
|
|
|
" mismatch for " + abs_path + ": " +
|
|
|
|
checksum_info);
|
|
|
|
}
|
2015-09-02 18:03:55 +02:00
|
|
|
}
|
|
|
|
}
|
2020-08-12 22:28:43 +02:00
|
|
|
|
2015-09-02 18:03:55 +02:00
|
|
|
return Status::OK();
|
|
|
|
}
|
|
|
|
|
2016-02-29 21:56:55 +01:00
|
|
|
Status BackupEngineImpl::CopyOrCreateFile(
|
|
|
|
const std::string& src, const std::string& dst, const std::string& contents,
|
2018-11-13 20:14:41 +01:00
|
|
|
Env* src_env, Env* dst_env, const EnvOptions& src_env_options, bool sync,
|
2020-08-12 22:28:43 +02:00
|
|
|
RateLimiter* rate_limiter, const std::string& backup_checksum_func_name,
|
|
|
|
uint64_t* size, std::string* checksum_hex, std::string* custom_checksum_hex,
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
uint64_t size_limit, std::function<void()> progress_callback) {
|
2016-02-29 21:56:55 +01:00
|
|
|
assert(src.empty() != contents.empty());
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
Status s;
|
2018-11-09 20:17:34 +01:00
|
|
|
std::unique_ptr<WritableFile> dst_file;
|
|
|
|
std::unique_ptr<SequentialFile> src_file;
|
2018-11-13 20:14:41 +01:00
|
|
|
EnvOptions dst_env_options;
|
|
|
|
dst_env_options.use_mmap_writes = false;
|
2017-05-23 03:40:41 +02:00
|
|
|
// TODO:(gzh) maybe use direct reads/writes here if possible
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
if (size != nullptr) {
|
|
|
|
*size = 0;
|
|
|
|
}
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
uint32_t checksum_value = 0;
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
|
2020-08-12 22:28:43 +02:00
|
|
|
// Get custom checksum function
|
|
|
|
std::unique_ptr<FileChecksumGenerator> checksum_func;
|
|
|
|
s = SetChecksumGenerator(backup_checksum_func_name, checksum_func);
|
|
|
|
if (!s.ok()) {
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
// Check if size limit is set. if not, set it to very big number
|
|
|
|
if (size_limit == 0) {
|
|
|
|
size_limit = std::numeric_limits<uint64_t>::max();
|
|
|
|
}
|
|
|
|
|
2018-11-13 20:14:41 +01:00
|
|
|
s = dst_env->NewWritableFile(dst, &dst_file, dst_env_options);
|
2016-02-29 21:56:55 +01:00
|
|
|
if (s.ok() && !src.empty()) {
|
2018-11-13 20:14:41 +01:00
|
|
|
s = src_env->NewSequentialFile(src, &src_file, src_env_options);
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
}
|
|
|
|
if (!s.ok()) {
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
Introduce a new storage specific Env API (#5761)
Summary:
The current Env API encompasses both storage/file operations, as well as OS related operations. Most of the APIs return a Status, which does not have enough metadata about an error, such as whether its retry-able or not, scope (i.e fault domain) of the error etc., that may be required in order to properly handle a storage error. The file APIs also do not provide enough control over the IO SLA, such as timeout, prioritization, hinting about placement and redundancy etc.
This PR separates out the file/storage APIs from Env into a new FileSystem class. The APIs are updated to return an IOStatus with metadata about the error, as well as to take an IOOptions structure as input in order to allow more control over the IO.
The user can set both ```options.env``` and ```options.file_system``` to specify that RocksDB should use the former for OS related operations and the latter for storage operations. Internally, a ```CompositeEnvWrapper``` has been introduced that inherits from ```Env``` and redirects individual methods to either an ```Env``` implementation or the ```FileSystem``` as appropriate. When options are sanitized during ```DB::Open```, ```options.env``` is replaced with a newly allocated ```CompositeEnvWrapper``` instance if both env and file_system have been specified. This way, the rest of the RocksDB code can continue to function as before.
This PR also ports PosixEnv to the new API by splitting it into two - PosixEnv and PosixFileSystem. PosixEnv is defined as a sub-class of CompositeEnvWrapper, and threading/time functions are overridden with Posix specific implementations in order to avoid an extra level of indirection.
The ```CompositeEnvWrapper``` translates ```IOStatus``` return code to ```Status```, and sets the severity to ```kSoftError``` if the io_status is retryable. The error handling code in RocksDB can then recover the DB automatically.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5761
Differential Revision: D18868376
Pulled By: anand1976
fbshipit-source-id: 39efe18a162ea746fabac6360ff529baba48486f
2019-12-13 23:47:08 +01:00
|
|
|
std::unique_ptr<WritableFileWriter> dest_writer(new WritableFileWriter(
|
|
|
|
NewLegacyWritableFileWrapper(std::move(dst_file)), dst, dst_env_options));
|
2018-11-09 20:17:34 +01:00
|
|
|
std::unique_ptr<SequentialFileReader> src_reader;
|
|
|
|
std::unique_ptr<char[]> buf;
|
2016-02-29 21:56:55 +01:00
|
|
|
if (!src.empty()) {
|
Introduce a new storage specific Env API (#5761)
Summary:
The current Env API encompasses both storage/file operations, as well as OS related operations. Most of the APIs return a Status, which does not have enough metadata about an error, such as whether its retry-able or not, scope (i.e fault domain) of the error etc., that may be required in order to properly handle a storage error. The file APIs also do not provide enough control over the IO SLA, such as timeout, prioritization, hinting about placement and redundancy etc.
This PR separates out the file/storage APIs from Env into a new FileSystem class. The APIs are updated to return an IOStatus with metadata about the error, as well as to take an IOOptions structure as input in order to allow more control over the IO.
The user can set both ```options.env``` and ```options.file_system``` to specify that RocksDB should use the former for OS related operations and the latter for storage operations. Internally, a ```CompositeEnvWrapper``` has been introduced that inherits from ```Env``` and redirects individual methods to either an ```Env``` implementation or the ```FileSystem``` as appropriate. When options are sanitized during ```DB::Open```, ```options.env``` is replaced with a newly allocated ```CompositeEnvWrapper``` instance if both env and file_system have been specified. This way, the rest of the RocksDB code can continue to function as before.
This PR also ports PosixEnv to the new API by splitting it into two - PosixEnv and PosixFileSystem. PosixEnv is defined as a sub-class of CompositeEnvWrapper, and threading/time functions are overridden with Posix specific implementations in order to avoid an extra level of indirection.
The ```CompositeEnvWrapper``` translates ```IOStatus``` return code to ```Status```, and sets the severity to ```kSoftError``` if the io_status is retryable. The error handling code in RocksDB can then recover the DB automatically.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5761
Differential Revision: D18868376
Pulled By: anand1976
fbshipit-source-id: 39efe18a162ea746fabac6360ff529baba48486f
2019-12-13 23:47:08 +01:00
|
|
|
src_reader.reset(new SequentialFileReader(
|
|
|
|
NewLegacySequentialFileWrapper(src_file), src));
|
2016-02-29 21:56:55 +01:00
|
|
|
buf.reset(new char[copy_file_buffer_size_]);
|
|
|
|
}
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
|
2016-02-29 21:56:55 +01:00
|
|
|
Slice data;
|
2015-09-10 03:49:15 +02:00
|
|
|
uint64_t processed_buffer_size = 0;
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
do {
|
2014-01-09 21:24:28 +01:00
|
|
|
if (stop_backup_.load(std::memory_order_acquire)) {
|
|
|
|
return Status::Incomplete("Backup stopped");
|
|
|
|
}
|
2016-02-29 21:56:55 +01:00
|
|
|
if (!src.empty()) {
|
|
|
|
size_t buffer_to_read = (copy_file_buffer_size_ < size_limit)
|
|
|
|
? copy_file_buffer_size_
|
2018-09-06 03:07:53 +02:00
|
|
|
: static_cast<size_t>(size_limit);
|
2016-02-29 21:56:55 +01:00
|
|
|
s = src_reader->Read(buffer_to_read, &data, buf.get());
|
|
|
|
processed_buffer_size += buffer_to_read;
|
|
|
|
} else {
|
|
|
|
data = contents;
|
|
|
|
}
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
size_limit -= data.size();
|
2020-07-03 03:13:31 +02:00
|
|
|
TEST_SYNC_POINT_CALLBACK(
|
|
|
|
"BackupEngineImpl::CopyOrCreateFile:CorruptionDuringBackup",
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
(src.length() > 4 && src.rfind(".sst") == src.length() - 4) ? &data
|
|
|
|
: nullptr);
|
2014-01-28 18:43:36 +01:00
|
|
|
|
|
|
|
if (!s.ok()) {
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
if (size != nullptr) {
|
|
|
|
*size += data.size();
|
|
|
|
}
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
if (checksum_hex != nullptr) {
|
|
|
|
checksum_value = crc32c::Extend(checksum_value, data.data(), data.size());
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
}
|
2020-08-12 22:28:43 +02:00
|
|
|
if (checksum_func != nullptr && custom_checksum_hex != nullptr) {
|
|
|
|
checksum_func->Update(data.data(), data.size());
|
|
|
|
}
|
|
|
|
|
Move rate_limiter, write buffering, most perf context instrumentation and most random kill out of Env
Summary: We want to keep Env a think layer for better portability. Less platform dependent codes should be moved out of Env. In this patch, I create a wrapper of file readers and writers, and put rate limiting, write buffering, as well as most perf context instrumentation and random kill out of Env. It will make it easier to maintain multiple Env in the future.
Test Plan: Run all existing unit tests.
Reviewers: anthony, kradhakrishnan, IslamAbdelRahman, yhchiang, igor
Reviewed By: igor
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D42321
2015-07-18 01:16:11 +02:00
|
|
|
s = dest_writer->Append(data);
|
2014-03-24 19:38:44 +01:00
|
|
|
if (rate_limiter != nullptr) {
|
2017-06-13 23:51:22 +02:00
|
|
|
rate_limiter->Request(data.size(), Env::IO_LOW, nullptr /* stats */,
|
|
|
|
RateLimiter::OpType::kWrite);
|
2014-03-24 19:38:44 +01:00
|
|
|
}
|
2015-09-10 03:49:15 +02:00
|
|
|
if (processed_buffer_size > options_.callback_trigger_interval_size) {
|
|
|
|
processed_buffer_size -= options_.callback_trigger_interval_size;
|
|
|
|
std::lock_guard<std::mutex> lock(byte_report_mutex_);
|
|
|
|
progress_callback();
|
|
|
|
}
|
2016-02-29 21:56:55 +01:00
|
|
|
} while (s.ok() && contents.empty() && data.size() > 0 && size_limit > 0);
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
if (checksum_hex != nullptr) {
|
2020-08-12 22:28:43 +02:00
|
|
|
// Convert uint32_t checksum to hex checksum
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
checksum_hex->assign(ChecksumInt32ToHex(checksum_value));
|
|
|
|
}
|
2020-08-12 22:28:43 +02:00
|
|
|
if (checksum_func != nullptr && custom_checksum_hex != nullptr) {
|
|
|
|
checksum_func->Finalize();
|
|
|
|
custom_checksum_hex->assign(ChecksumStrToHex(checksum_func->GetChecksum()));
|
|
|
|
}
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
if (s.ok() && sync) {
|
Move rate_limiter, write buffering, most perf context instrumentation and most random kill out of Env
Summary: We want to keep Env a think layer for better portability. Less platform dependent codes should be moved out of Env. In this patch, I create a wrapper of file readers and writers, and put rate limiting, write buffering, as well as most perf context instrumentation and random kill out of Env. It will make it easier to maintain multiple Env in the future.
Test Plan: Run all existing unit tests.
Reviewers: anthony, kradhakrishnan, IslamAbdelRahman, yhchiang, igor
Reviewed By: igor
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D42321
2015-07-18 01:16:11 +02:00
|
|
|
s = dest_writer->Sync(false);
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
}
|
2016-06-30 00:33:27 +02:00
|
|
|
if (s.ok()) {
|
|
|
|
s = dest_writer->Close();
|
|
|
|
}
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
2016-02-29 21:56:55 +01:00
|
|
|
// fname will always start with "/"
|
2015-07-02 20:35:51 +02:00
|
|
|
Status BackupEngineImpl::AddBackupFileWorkItem(
|
2015-09-10 03:49:15 +02:00
|
|
|
std::unordered_set<std::string>& live_dst_paths,
|
2016-02-29 21:56:55 +01:00
|
|
|
std::vector<BackupAfterCopyOrCreateWorkItem>& backup_items_to_finish,
|
2015-09-10 03:49:15 +02:00
|
|
|
BackupID backup_id, bool shared, const std::string& src_dir,
|
2018-11-13 20:14:41 +01:00
|
|
|
const std::string& fname, const EnvOptions& src_env_options,
|
|
|
|
RateLimiter* rate_limiter, uint64_t size_bytes, uint64_t size_limit,
|
|
|
|
bool shared_checksum, std::function<void()> progress_callback,
|
2020-07-03 03:13:31 +02:00
|
|
|
const std::string& contents, const std::string& src_checksum_func_name,
|
|
|
|
const std::string& src_checksum_str) {
|
2016-02-29 21:56:55 +01:00
|
|
|
assert(!fname.empty() && fname[0] == '/');
|
|
|
|
assert(contents.empty() != src_dir.empty());
|
|
|
|
|
|
|
|
std::string dst_relative = fname.substr(1);
|
2014-01-09 21:24:28 +01:00
|
|
|
std::string dst_relative_tmp;
|
2014-05-03 02:08:55 +02:00
|
|
|
Status s;
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
std::string checksum_hex;
|
2020-08-12 22:28:43 +02:00
|
|
|
std::string custom_checksum_hex;
|
|
|
|
// The function name of backup checksum function.
|
|
|
|
std::string backup_checksum_func_name = kDefaultBackupFileChecksumFuncName;
|
2020-06-25 04:30:15 +02:00
|
|
|
std::string db_id;
|
|
|
|
std::string db_session_id;
|
2020-08-14 22:31:56 +02:00
|
|
|
// whether a default or custom checksum for a table file is available
|
2020-07-03 03:13:31 +02:00
|
|
|
bool has_checksum = false;
|
2014-05-03 02:08:55 +02:00
|
|
|
|
2020-08-12 22:28:43 +02:00
|
|
|
// Set up the custom checksum function.
|
|
|
|
// A nullptr checksum_func indicates the default backup checksum function
|
|
|
|
// will be used. If checksum_func is not nullptr, then both the default
|
|
|
|
// backup checksum function and checksum_func will be used.
|
|
|
|
std::unique_ptr<FileChecksumGenerator> checksum_func;
|
|
|
|
if (src_checksum_func_name != kUnknownFileChecksumFuncName) {
|
|
|
|
// DB files have checksum functions
|
|
|
|
s = SetChecksumGenerator(src_checksum_func_name, checksum_func);
|
|
|
|
if (!s.ok()) {
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
if (checksum_func != nullptr) {
|
|
|
|
backup_checksum_func_name = checksum_func->Name();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
// Whenever the db checksum function name matches the backup engine custom
|
|
|
|
// checksum function name, we will compare the checksum values after copying.
|
|
|
|
// Note that only table files may have a known checksum name passed in.
|
2020-07-03 03:13:31 +02:00
|
|
|
//
|
2020-08-12 22:28:43 +02:00
|
|
|
// If the checksum function names do not match and db session id is not
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
// available, we will calculate the checksum *before* copying in two cases
|
|
|
|
// (we always calcuate checksums when copying or creating for any file types):
|
2020-07-03 03:13:31 +02:00
|
|
|
// a) share_files_with_checksum is true and file type is table;
|
|
|
|
// b) share_table_files is true and the file exists already.
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
//
|
2020-08-12 22:28:43 +02:00
|
|
|
// Step 0: Check if a known checksum function name is passed in
|
|
|
|
if (IsSameChecksumFunc(backup_checksum_func_name, src_checksum_func_name)) {
|
2020-07-03 03:13:31 +02:00
|
|
|
if (src_checksum_str == kUnknownFileChecksum) {
|
2020-07-04 00:38:35 +02:00
|
|
|
return Status::Aborted("Unknown checksum value for " + fname);
|
2020-07-03 03:13:31 +02:00
|
|
|
}
|
2020-08-12 22:28:43 +02:00
|
|
|
if (checksum_func == nullptr) {
|
|
|
|
checksum_hex = ChecksumStrToHex(src_checksum_str);
|
|
|
|
} else {
|
|
|
|
custom_checksum_hex = ChecksumStrToHex(src_checksum_str);
|
|
|
|
}
|
2020-07-03 03:13:31 +02:00
|
|
|
has_checksum = true;
|
|
|
|
}
|
|
|
|
|
|
|
|
// Step 1: Prepare the relative path to destination
|
|
|
|
if (shared && shared_checksum) {
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
if (GetTableNamingOption() == kOptionalChecksumAndDbSessionId) {
|
2020-06-25 04:30:15 +02:00
|
|
|
// Prepare db_session_id to add to the file name
|
|
|
|
// Ignore the returned status
|
|
|
|
// In the failed cases, db_id and db_session_id will be empty
|
2020-07-09 17:36:41 +02:00
|
|
|
GetFileDbIdentities(db_env_, src_env_options, src_dir + fname, &db_id,
|
|
|
|
&db_session_id);
|
2020-06-25 04:30:15 +02:00
|
|
|
}
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
// Calculate checksum if checksum and db session id are not available.
|
|
|
|
// If db session id is available, we will not calculate the checksum
|
|
|
|
// since the session id should suffice to avoid file name collision in
|
|
|
|
// the shared_checksum directory.
|
|
|
|
if (!has_checksum && db_session_id.empty()) {
|
|
|
|
s = CalculateChecksum(src_dir + fname, db_env_, src_env_options,
|
2020-08-12 22:28:43 +02:00
|
|
|
size_limit, &checksum_hex, checksum_func,
|
|
|
|
&custom_checksum_hex);
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
if (!s.ok()) {
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
has_checksum = true;
|
|
|
|
}
|
2016-03-02 04:33:33 +01:00
|
|
|
if (size_bytes == port::kMaxUint64) {
|
|
|
|
return Status::NotFound("File missing: " + src_dir + fname);
|
2014-05-03 02:08:55 +02:00
|
|
|
}
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
// dst_relative depends on the following conditions:
|
|
|
|
// 1) the naming scheme is kOptionalChecksumAndDbSessionId,
|
|
|
|
// 2) db_session_id is not empty,
|
|
|
|
// 3) checksum is available in the DB manifest.
|
|
|
|
// If 1,2,3) are satisfied, then dst_relative will be of the form:
|
|
|
|
// shared_checksum/<file_number>_<checksum>_<db_session_id>.sst
|
|
|
|
// If 1,2) are satisfied, then dst_relative will be of the form:
|
|
|
|
// shared_checksum/<file_number>_<db_session_id>.sst
|
|
|
|
// Otherwise, dst_relative is of the form
|
|
|
|
// shared_checksum/<file_number>_<checksum>_<size>.sst
|
2020-08-12 22:28:43 +02:00
|
|
|
//
|
|
|
|
// Also, we display custom checksums in the name if possible.
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
dst_relative = GetSharedFileWithChecksum(
|
2020-08-12 22:28:43 +02:00
|
|
|
dst_relative, has_checksum,
|
|
|
|
checksum_func == nullptr || !UseSessionId(db_session_id)
|
|
|
|
? checksum_hex
|
|
|
|
: custom_checksum_hex,
|
|
|
|
size_bytes, db_session_id);
|
2014-05-03 02:08:55 +02:00
|
|
|
dst_relative_tmp = GetSharedFileWithChecksumRel(dst_relative, true);
|
|
|
|
dst_relative = GetSharedFileWithChecksumRel(dst_relative, false);
|
|
|
|
} else if (shared) {
|
2014-01-09 21:24:28 +01:00
|
|
|
dst_relative_tmp = GetSharedFileRel(dst_relative, true);
|
|
|
|
dst_relative = GetSharedFileRel(dst_relative, false);
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
} else {
|
2014-01-09 21:24:28 +01:00
|
|
|
dst_relative = GetPrivateFileRel(backup_id, false, dst_relative);
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
}
|
2018-04-21 02:23:34 +02:00
|
|
|
|
|
|
|
// We copy into `temp_dest_path` and, once finished, rename it to
|
|
|
|
// `final_dest_path`. This allows files to atomically appear at
|
|
|
|
// `final_dest_path`. We can copy directly to the final path when atomicity
|
|
|
|
// is unnecessary, like for files in private backup directories.
|
|
|
|
const std::string* copy_dest_path;
|
|
|
|
std::string temp_dest_path;
|
|
|
|
std::string final_dest_path = GetAbsolutePath(dst_relative);
|
|
|
|
if (!dst_relative_tmp.empty()) {
|
|
|
|
temp_dest_path = GetAbsolutePath(dst_relative_tmp);
|
|
|
|
copy_dest_path = &temp_dest_path;
|
|
|
|
} else {
|
|
|
|
copy_dest_path = &final_dest_path;
|
|
|
|
}
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
|
2020-06-25 04:30:15 +02:00
|
|
|
// Step 2: Determine whether to copy or not
|
2015-07-02 20:35:51 +02:00
|
|
|
// if it's shared, we also need to check if it exists -- if it does, no need
|
|
|
|
// to copy it again.
|
Fix BackupEngine
Summary:
In D28521 we removed GarbageCollect() from BackupEngine's constructor. The reason was that opening BackupEngine on HDFS was very slow and in most cases we didn't have any garbage. We allowed the user to call GarbageCollect() when it detects some garbage files in his backup directory.
Unfortunately, this left us vulnerable to an interesting issue. Let's say we started a backup and copied files {1, 3} but the backup failed. On another host, we restore DB from backup and generate {1, 3, 5}. Since {1, 3} is already there, we will not overwrite. However, these files might be from a different database so their contents might be different. See internal task t6781803 for more info.
Now, when we're copying files and we discover a file already there, we check:
1. if the file is not referenced from any backups, we overwrite the file.
2. if the file is referenced from other backups AND the checksums don't match, we fail the backup. This will only happen if user is using a single backup directory for backing up two different databases.
3. if the file is referenced from other backups AND the checksums match, it's all good. We skip the copy and go copy the next file.
Test Plan: Added new test to backupable_db_test. The test fails before this patch.
Reviewers: sdong, rven, yhchiang
Reviewed By: yhchiang
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D37599
2015-05-08 02:39:19 +02:00
|
|
|
bool need_to_copy = true;
|
2018-04-21 02:23:34 +02:00
|
|
|
// true if final_dest_path is the same path as another live file
|
2015-07-02 20:35:51 +02:00
|
|
|
const bool same_path =
|
2018-04-21 02:23:34 +02:00
|
|
|
live_dst_paths.find(final_dest_path) != live_dst_paths.end();
|
2015-07-21 02:20:40 +02:00
|
|
|
|
|
|
|
bool file_exists = false;
|
|
|
|
if (shared && !same_path) {
|
2020-06-25 04:30:15 +02:00
|
|
|
// Should be in shared directory but not a live path, check existence in
|
|
|
|
// shared directory
|
2018-04-21 02:23:34 +02:00
|
|
|
Status exist = backup_env_->FileExists(final_dest_path);
|
2015-07-21 02:20:40 +02:00
|
|
|
if (exist.ok()) {
|
|
|
|
file_exists = true;
|
|
|
|
} else if (exist.IsNotFound()) {
|
|
|
|
file_exists = false;
|
|
|
|
} else {
|
|
|
|
assert(s.IsIOError());
|
|
|
|
return exist;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2016-02-29 21:56:55 +01:00
|
|
|
if (!contents.empty()) {
|
|
|
|
need_to_copy = false;
|
|
|
|
} else if (shared && (same_path || file_exists)) {
|
Fix BackupEngine
Summary:
In D28521 we removed GarbageCollect() from BackupEngine's constructor. The reason was that opening BackupEngine on HDFS was very slow and in most cases we didn't have any garbage. We allowed the user to call GarbageCollect() when it detects some garbage files in his backup directory.
Unfortunately, this left us vulnerable to an interesting issue. Let's say we started a backup and copied files {1, 3} but the backup failed. On another host, we restore DB from backup and generate {1, 3, 5}. Since {1, 3} is already there, we will not overwrite. However, these files might be from a different database so their contents might be different. See internal task t6781803 for more info.
Now, when we're copying files and we discover a file already there, we check:
1. if the file is not referenced from any backups, we overwrite the file.
2. if the file is referenced from other backups AND the checksums don't match, we fail the backup. This will only happen if user is using a single backup directory for backing up two different databases.
3. if the file is referenced from other backups AND the checksums match, it's all good. We skip the copy and go copy the next file.
Test Plan: Added new test to backupable_db_test. The test fails before this patch.
Reviewers: sdong, rven, yhchiang
Reviewed By: yhchiang
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D37599
2015-05-08 02:39:19 +02:00
|
|
|
need_to_copy = false;
|
2014-05-03 02:08:55 +02:00
|
|
|
if (shared_checksum) {
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
if (backuped_file_infos_.find(dst_relative) ==
|
|
|
|
backuped_file_infos_.end() &&
|
|
|
|
!same_path) {
|
|
|
|
// file exists but not referenced
|
|
|
|
ROCKS_LOG_INFO(
|
|
|
|
options_.info_log,
|
|
|
|
"%s already present, but not referenced by any backup. We will "
|
|
|
|
"overwrite the file.",
|
|
|
|
fname.c_str());
|
|
|
|
need_to_copy = true;
|
|
|
|
backup_env_->DeleteFile(final_dest_path);
|
2020-06-25 04:30:15 +02:00
|
|
|
} else {
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
// file exists and referenced
|
2020-08-14 22:31:56 +02:00
|
|
|
if (!has_checksum || checksum_hex.empty()) {
|
|
|
|
// Either both checksum_hex and custom_checksum_hex need recalculating
|
|
|
|
// or only checksum_hex needs recalculating
|
|
|
|
s = CalculateChecksum(
|
|
|
|
src_dir + fname, db_env_, src_env_options, size_limit,
|
|
|
|
&checksum_hex, checksum_func,
|
|
|
|
checksum_hex.empty() ? nullptr : &custom_checksum_hex);
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
if (!s.ok()) {
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
has_checksum = true;
|
|
|
|
}
|
|
|
|
if (UseSessionId(db_session_id)) {
|
|
|
|
ROCKS_LOG_INFO(options_.info_log,
|
|
|
|
"%s already present, with checksum %s, size %" PRIu64
|
|
|
|
" and DB session identity %s",
|
|
|
|
fname.c_str(), checksum_hex.c_str(), size_bytes,
|
|
|
|
db_session_id.c_str());
|
|
|
|
} else {
|
|
|
|
ROCKS_LOG_INFO(
|
|
|
|
options_.info_log,
|
|
|
|
"%s already present, with checksum %s and size %" PRIu64,
|
|
|
|
fname.c_str(), checksum_hex.c_str(), size_bytes);
|
|
|
|
}
|
2020-06-25 04:30:15 +02:00
|
|
|
}
|
2020-08-12 22:28:43 +02:00
|
|
|
if (checksum_func != nullptr) {
|
|
|
|
ROCKS_LOG_INFO(options_.info_log, "%s checksum is %s",
|
|
|
|
backup_checksum_func_name.c_str(),
|
|
|
|
custom_checksum_hex.c_str());
|
|
|
|
}
|
Fix BackupEngine
Summary:
In D28521 we removed GarbageCollect() from BackupEngine's constructor. The reason was that opening BackupEngine on HDFS was very slow and in most cases we didn't have any garbage. We allowed the user to call GarbageCollect() when it detects some garbage files in his backup directory.
Unfortunately, this left us vulnerable to an interesting issue. Let's say we started a backup and copied files {1, 3} but the backup failed. On another host, we restore DB from backup and generate {1, 3, 5}. Since {1, 3} is already there, we will not overwrite. However, these files might be from a different database so their contents might be different. See internal task t6781803 for more info.
Now, when we're copying files and we discover a file already there, we check:
1. if the file is not referenced from any backups, we overwrite the file.
2. if the file is referenced from other backups AND the checksums don't match, we fail the backup. This will only happen if user is using a single backup directory for backing up two different databases.
3. if the file is referenced from other backups AND the checksums match, it's all good. We skip the copy and go copy the next file.
Test Plan: Added new test to backupable_db_test. The test fails before this patch.
Reviewers: sdong, rven, yhchiang
Reviewed By: yhchiang
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D37599
2015-05-08 02:39:19 +02:00
|
|
|
} else if (backuped_file_infos_.find(dst_relative) ==
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
backuped_file_infos_.end() &&
|
|
|
|
!same_path) {
|
Fix BackupEngine
Summary:
In D28521 we removed GarbageCollect() from BackupEngine's constructor. The reason was that opening BackupEngine on HDFS was very slow and in most cases we didn't have any garbage. We allowed the user to call GarbageCollect() when it detects some garbage files in his backup directory.
Unfortunately, this left us vulnerable to an interesting issue. Let's say we started a backup and copied files {1, 3} but the backup failed. On another host, we restore DB from backup and generate {1, 3, 5}. Since {1, 3} is already there, we will not overwrite. However, these files might be from a different database so their contents might be different. See internal task t6781803 for more info.
Now, when we're copying files and we discover a file already there, we check:
1. if the file is not referenced from any backups, we overwrite the file.
2. if the file is referenced from other backups AND the checksums don't match, we fail the backup. This will only happen if user is using a single backup directory for backing up two different databases.
3. if the file is referenced from other backups AND the checksums match, it's all good. We skip the copy and go copy the next file.
Test Plan: Added new test to backupable_db_test. The test fails before this patch.
Reviewers: sdong, rven, yhchiang
Reviewed By: yhchiang
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D37599
2015-05-08 02:39:19 +02:00
|
|
|
// file already exists, but it's not referenced by any backup. overwrite
|
|
|
|
// the file
|
2017-03-16 03:22:52 +01:00
|
|
|
ROCKS_LOG_INFO(
|
|
|
|
options_.info_log,
|
Fix BackupEngine
Summary:
In D28521 we removed GarbageCollect() from BackupEngine's constructor. The reason was that opening BackupEngine on HDFS was very slow and in most cases we didn't have any garbage. We allowed the user to call GarbageCollect() when it detects some garbage files in his backup directory.
Unfortunately, this left us vulnerable to an interesting issue. Let's say we started a backup and copied files {1, 3} but the backup failed. On another host, we restore DB from backup and generate {1, 3, 5}. Since {1, 3} is already there, we will not overwrite. However, these files might be from a different database so their contents might be different. See internal task t6781803 for more info.
Now, when we're copying files and we discover a file already there, we check:
1. if the file is not referenced from any backups, we overwrite the file.
2. if the file is referenced from other backups AND the checksums don't match, we fail the backup. This will only happen if user is using a single backup directory for backing up two different databases.
3. if the file is referenced from other backups AND the checksums match, it's all good. We skip the copy and go copy the next file.
Test Plan: Added new test to backupable_db_test. The test fails before this patch.
Reviewers: sdong, rven, yhchiang
Reviewed By: yhchiang
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D37599
2015-05-08 02:39:19 +02:00
|
|
|
"%s already present, but not referenced by any backup. We will "
|
|
|
|
"overwrite the file.",
|
2016-02-29 21:56:55 +01:00
|
|
|
fname.c_str());
|
Fix BackupEngine
Summary:
In D28521 we removed GarbageCollect() from BackupEngine's constructor. The reason was that opening BackupEngine on HDFS was very slow and in most cases we didn't have any garbage. We allowed the user to call GarbageCollect() when it detects some garbage files in his backup directory.
Unfortunately, this left us vulnerable to an interesting issue. Let's say we started a backup and copied files {1, 3} but the backup failed. On another host, we restore DB from backup and generate {1, 3, 5}. Since {1, 3} is already there, we will not overwrite. However, these files might be from a different database so their contents might be different. See internal task t6781803 for more info.
Now, when we're copying files and we discover a file already there, we check:
1. if the file is not referenced from any backups, we overwrite the file.
2. if the file is referenced from other backups AND the checksums don't match, we fail the backup. This will only happen if user is using a single backup directory for backing up two different databases.
3. if the file is referenced from other backups AND the checksums match, it's all good. We skip the copy and go copy the next file.
Test Plan: Added new test to backupable_db_test. The test fails before this patch.
Reviewers: sdong, rven, yhchiang
Reviewed By: yhchiang
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D37599
2015-05-08 02:39:19 +02:00
|
|
|
need_to_copy = true;
|
2018-04-21 02:23:34 +02:00
|
|
|
backup_env_->DeleteFile(final_dest_path);
|
2014-05-03 02:08:55 +02:00
|
|
|
} else {
|
Fix BackupEngine
Summary:
In D28521 we removed GarbageCollect() from BackupEngine's constructor. The reason was that opening BackupEngine on HDFS was very slow and in most cases we didn't have any garbage. We allowed the user to call GarbageCollect() when it detects some garbage files in his backup directory.
Unfortunately, this left us vulnerable to an interesting issue. Let's say we started a backup and copied files {1, 3} but the backup failed. On another host, we restore DB from backup and generate {1, 3, 5}. Since {1, 3} is already there, we will not overwrite. However, these files might be from a different database so their contents might be different. See internal task t6781803 for more info.
Now, when we're copying files and we discover a file already there, we check:
1. if the file is not referenced from any backups, we overwrite the file.
2. if the file is referenced from other backups AND the checksums don't match, we fail the backup. This will only happen if user is using a single backup directory for backing up two different databases.
3. if the file is referenced from other backups AND the checksums match, it's all good. We skip the copy and go copy the next file.
Test Plan: Added new test to backupable_db_test. The test fails before this patch.
Reviewers: sdong, rven, yhchiang
Reviewed By: yhchiang
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D37599
2015-05-08 02:39:19 +02:00
|
|
|
// the file is present and referenced by a backup
|
2017-03-16 03:22:52 +01:00
|
|
|
ROCKS_LOG_INFO(options_.info_log,
|
|
|
|
"%s already present, calculate checksum", fname.c_str());
|
2020-08-14 22:31:56 +02:00
|
|
|
if (!has_checksum || checksum_hex.empty()) {
|
|
|
|
// Either both checksum_hex and custom_checksum_hex need recalculating
|
|
|
|
// or only checksum_hex needs recalculating
|
|
|
|
s = CalculateChecksum(
|
|
|
|
src_dir + fname, db_env_, src_env_options, size_limit,
|
|
|
|
&checksum_hex, checksum_func,
|
|
|
|
checksum_hex.empty() ? nullptr : &custom_checksum_hex);
|
2020-07-03 03:13:31 +02:00
|
|
|
if (!s.ok()) {
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
has_checksum = true;
|
2020-06-25 04:30:15 +02:00
|
|
|
}
|
2014-05-03 02:08:55 +02:00
|
|
|
}
|
Fix BackupEngine
Summary:
In D28521 we removed GarbageCollect() from BackupEngine's constructor. The reason was that opening BackupEngine on HDFS was very slow and in most cases we didn't have any garbage. We allowed the user to call GarbageCollect() when it detects some garbage files in his backup directory.
Unfortunately, this left us vulnerable to an interesting issue. Let's say we started a backup and copied files {1, 3} but the backup failed. On another host, we restore DB from backup and generate {1, 3, 5}. Since {1, 3} is already there, we will not overwrite. However, these files might be from a different database so their contents might be different. See internal task t6781803 for more info.
Now, when we're copying files and we discover a file already there, we check:
1. if the file is not referenced from any backups, we overwrite the file.
2. if the file is referenced from other backups AND the checksums don't match, we fail the backup. This will only happen if user is using a single backup directory for backing up two different databases.
3. if the file is referenced from other backups AND the checksums match, it's all good. We skip the copy and go copy the next file.
Test Plan: Added new test to backupable_db_test. The test fails before this patch.
Reviewers: sdong, rven, yhchiang
Reviewed By: yhchiang
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D37599
2015-05-08 02:39:19 +02:00
|
|
|
}
|
2018-04-21 02:23:34 +02:00
|
|
|
live_dst_paths.insert(final_dest_path);
|
2015-07-02 20:35:51 +02:00
|
|
|
|
2020-06-25 04:30:15 +02:00
|
|
|
// Step 3: Add work item
|
2016-02-29 21:56:55 +01:00
|
|
|
if (!contents.empty() || need_to_copy) {
|
2017-03-16 03:22:52 +01:00
|
|
|
ROCKS_LOG_INFO(options_.info_log, "Copying %s to %s", fname.c_str(),
|
2018-04-21 02:23:34 +02:00
|
|
|
copy_dest_path->c_str());
|
2016-02-29 21:56:55 +01:00
|
|
|
CopyOrCreateWorkItem copy_or_create_work_item(
|
2018-04-21 02:23:34 +02:00
|
|
|
src_dir.empty() ? "" : src_dir + fname, *copy_dest_path, contents,
|
2018-11-13 20:14:41 +01:00
|
|
|
db_env_, backup_env_, src_env_options, options_.sync, rate_limiter,
|
2020-07-03 03:13:31 +02:00
|
|
|
size_limit, progress_callback, has_checksum, src_checksum_func_name,
|
2020-08-12 22:28:43 +02:00
|
|
|
checksum_func == nullptr ? checksum_hex : custom_checksum_hex,
|
|
|
|
backup_checksum_func_name, db_id, db_session_id);
|
2016-02-29 21:56:55 +01:00
|
|
|
BackupAfterCopyOrCreateWorkItem after_copy_or_create_work_item(
|
|
|
|
copy_or_create_work_item.result.get_future(), shared, need_to_copy,
|
2018-04-21 02:23:34 +02:00
|
|
|
backup_env_, temp_dest_path, final_dest_path, dst_relative);
|
2016-02-29 21:56:55 +01:00
|
|
|
files_to_copy_or_create_.write(std::move(copy_or_create_work_item));
|
|
|
|
backup_items_to_finish.push_back(std::move(after_copy_or_create_work_item));
|
2015-07-02 20:35:51 +02:00
|
|
|
} else {
|
2016-02-29 21:56:55 +01:00
|
|
|
std::promise<CopyOrCreateResult> promise_result;
|
|
|
|
BackupAfterCopyOrCreateWorkItem after_copy_or_create_work_item(
|
|
|
|
promise_result.get_future(), shared, need_to_copy, backup_env_,
|
2018-04-21 02:23:34 +02:00
|
|
|
temp_dest_path, final_dest_path, dst_relative);
|
2016-02-29 21:56:55 +01:00
|
|
|
backup_items_to_finish.push_back(std::move(after_copy_or_create_work_item));
|
|
|
|
CopyOrCreateResult result;
|
2015-07-02 20:35:51 +02:00
|
|
|
result.status = s;
|
2016-03-02 04:33:33 +01:00
|
|
|
result.size = size_bytes;
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
result.checksum_hex = std::move(checksum_hex);
|
2020-08-12 22:28:43 +02:00
|
|
|
result.custom_checksum_hex = std::move(custom_checksum_hex);
|
|
|
|
result.checksum_func_name = std::move(backup_checksum_func_name);
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
result.db_id = std::move(db_id);
|
|
|
|
result.db_session_id = std::move(db_session_id);
|
2015-07-02 20:35:51 +02:00
|
|
|
promise_result.set_value(std::move(result));
|
2014-01-28 18:43:36 +01:00
|
|
|
}
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
2020-08-12 22:28:43 +02:00
|
|
|
Status BackupEngineImpl::CalculateChecksum(
|
|
|
|
const std::string& src, Env* src_env, const EnvOptions& src_env_options,
|
|
|
|
uint64_t size_limit, std::string* checksum_hex,
|
|
|
|
const std::unique_ptr<FileChecksumGenerator>& checksum_func,
|
|
|
|
std::string* custom_checksum_hex) {
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
if (checksum_hex == nullptr) {
|
2020-08-12 22:28:43 +02:00
|
|
|
return Status::InvalidArgument("Checksum pointer is null");
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
}
|
|
|
|
uint32_t checksum_value = 0;
|
2020-08-12 22:28:43 +02:00
|
|
|
|
2014-01-28 18:43:36 +01:00
|
|
|
if (size_limit == 0) {
|
|
|
|
size_limit = std::numeric_limits<uint64_t>::max();
|
|
|
|
}
|
|
|
|
|
|
|
|
std::unique_ptr<SequentialFile> src_file;
|
2018-11-13 20:14:41 +01:00
|
|
|
Status s = src_env->NewSequentialFile(src, &src_file, src_env_options);
|
2014-01-28 18:43:36 +01:00
|
|
|
if (!s.ok()) {
|
|
|
|
return s;
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
}
|
2014-01-28 18:43:36 +01:00
|
|
|
|
2018-11-09 20:17:34 +01:00
|
|
|
std::unique_ptr<SequentialFileReader> src_reader(
|
Introduce a new storage specific Env API (#5761)
Summary:
The current Env API encompasses both storage/file operations, as well as OS related operations. Most of the APIs return a Status, which does not have enough metadata about an error, such as whether its retry-able or not, scope (i.e fault domain) of the error etc., that may be required in order to properly handle a storage error. The file APIs also do not provide enough control over the IO SLA, such as timeout, prioritization, hinting about placement and redundancy etc.
This PR separates out the file/storage APIs from Env into a new FileSystem class. The APIs are updated to return an IOStatus with metadata about the error, as well as to take an IOOptions structure as input in order to allow more control over the IO.
The user can set both ```options.env``` and ```options.file_system``` to specify that RocksDB should use the former for OS related operations and the latter for storage operations. Internally, a ```CompositeEnvWrapper``` has been introduced that inherits from ```Env``` and redirects individual methods to either an ```Env``` implementation or the ```FileSystem``` as appropriate. When options are sanitized during ```DB::Open```, ```options.env``` is replaced with a newly allocated ```CompositeEnvWrapper``` instance if both env and file_system have been specified. This way, the rest of the RocksDB code can continue to function as before.
This PR also ports PosixEnv to the new API by splitting it into two - PosixEnv and PosixFileSystem. PosixEnv is defined as a sub-class of CompositeEnvWrapper, and threading/time functions are overridden with Posix specific implementations in order to avoid an extra level of indirection.
The ```CompositeEnvWrapper``` translates ```IOStatus``` return code to ```Status```, and sets the severity to ```kSoftError``` if the io_status is retryable. The error handling code in RocksDB can then recover the DB automatically.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5761
Differential Revision: D18868376
Pulled By: anand1976
fbshipit-source-id: 39efe18a162ea746fabac6360ff529baba48486f
2019-12-13 23:47:08 +01:00
|
|
|
new SequentialFileReader(NewLegacySequentialFileWrapper(src_file), src));
|
2014-01-28 18:43:36 +01:00
|
|
|
std::unique_ptr<char[]> buf(new char[copy_file_buffer_size_]);
|
|
|
|
Slice data;
|
|
|
|
|
|
|
|
do {
|
|
|
|
if (stop_backup_.load(std::memory_order_acquire)) {
|
|
|
|
return Status::Incomplete("Backup stopped");
|
|
|
|
}
|
|
|
|
size_t buffer_to_read = (copy_file_buffer_size_ < size_limit) ?
|
2018-09-06 03:07:53 +02:00
|
|
|
copy_file_buffer_size_ : static_cast<size_t>(size_limit);
|
Move rate_limiter, write buffering, most perf context instrumentation and most random kill out of Env
Summary: We want to keep Env a think layer for better portability. Less platform dependent codes should be moved out of Env. In this patch, I create a wrapper of file readers and writers, and put rate limiting, write buffering, as well as most perf context instrumentation and random kill out of Env. It will make it easier to maintain multiple Env in the future.
Test Plan: Run all existing unit tests.
Reviewers: anthony, kradhakrishnan, IslamAbdelRahman, yhchiang, igor
Reviewed By: igor
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D42321
2015-07-18 01:16:11 +02:00
|
|
|
s = src_reader->Read(buffer_to_read, &data, buf.get());
|
2014-01-28 18:43:36 +01:00
|
|
|
|
|
|
|
if (!s.ok()) {
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
|
|
|
size_limit -= data.size();
|
2020-08-12 22:28:43 +02:00
|
|
|
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
checksum_value = crc32c::Extend(checksum_value, data.data(), data.size());
|
2020-08-12 22:28:43 +02:00
|
|
|
if (checksum_func != nullptr && custom_checksum_hex != nullptr) {
|
|
|
|
checksum_func->Update(data.data(), data.size());
|
|
|
|
}
|
2014-01-28 18:43:36 +01:00
|
|
|
} while (data.size() > 0 && size_limit > 0);
|
|
|
|
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
checksum_hex->assign(ChecksumInt32ToHex(checksum_value));
|
2020-08-12 22:28:43 +02:00
|
|
|
if (checksum_func != nullptr && custom_checksum_hex != nullptr) {
|
|
|
|
checksum_func->Finalize();
|
|
|
|
custom_checksum_hex->assign(ChecksumStrToHex(checksum_func->GetChecksum()));
|
|
|
|
}
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
2020-06-25 04:30:15 +02:00
|
|
|
Status BackupEngineImpl::GetFileDbIdentities(Env* src_env,
|
2020-07-09 17:36:41 +02:00
|
|
|
const EnvOptions& src_env_options,
|
2020-06-25 04:30:15 +02:00
|
|
|
const std::string& file_path,
|
|
|
|
std::string* db_id,
|
|
|
|
std::string* db_session_id) {
|
2020-07-02 02:26:39 +02:00
|
|
|
assert(db_id != nullptr || db_session_id != nullptr);
|
2020-07-09 17:36:41 +02:00
|
|
|
|
2020-07-01 03:45:36 +02:00
|
|
|
Options options;
|
|
|
|
options.env = src_env;
|
2020-07-09 17:36:41 +02:00
|
|
|
SstFileDumper sst_reader(options, file_path,
|
2020-06-25 04:30:15 +02:00
|
|
|
2 * 1024 * 1024
|
|
|
|
/* readahead_size */,
|
|
|
|
false /* verify_checksum */, false /* output_hex */,
|
2020-07-09 17:36:41 +02:00
|
|
|
false /* decode_blob_index */, src_env_options,
|
|
|
|
true /* silent */);
|
2020-06-25 04:30:15 +02:00
|
|
|
|
|
|
|
const TableProperties* table_properties = nullptr;
|
|
|
|
std::shared_ptr<const TableProperties> tp;
|
|
|
|
Status s = sst_reader.getStatus();
|
|
|
|
|
|
|
|
if (s.ok()) {
|
|
|
|
// Try to get table properties from the table reader of sst_reader
|
|
|
|
if (!sst_reader.ReadTableProperties(&tp).ok()) {
|
|
|
|
// Try to use table properites from the initialization of sst_reader
|
|
|
|
table_properties = sst_reader.GetInitTableProperties();
|
|
|
|
} else {
|
|
|
|
table_properties = tp.get();
|
|
|
|
}
|
|
|
|
} else {
|
2020-07-09 17:36:41 +02:00
|
|
|
ROCKS_LOG_INFO(options_.info_log, "Failed to read %s: %s",
|
|
|
|
file_path.c_str(), s.ToString().c_str());
|
2020-06-25 04:30:15 +02:00
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (table_properties != nullptr) {
|
2020-07-02 02:26:39 +02:00
|
|
|
if (db_id != nullptr) {
|
|
|
|
db_id->assign(table_properties->db_id);
|
|
|
|
}
|
|
|
|
if (db_session_id != nullptr) {
|
|
|
|
db_session_id->assign(table_properties->db_session_id);
|
|
|
|
if (db_session_id->empty()) {
|
2020-07-09 17:36:41 +02:00
|
|
|
s = Status::NotFound("DB session identity not found in " + file_path);
|
|
|
|
ROCKS_LOG_INFO(options_.info_log, "%s", s.ToString().c_str());
|
|
|
|
return s;
|
2020-07-02 02:26:39 +02:00
|
|
|
}
|
2020-06-25 04:30:15 +02:00
|
|
|
}
|
|
|
|
return Status::OK();
|
|
|
|
} else {
|
2020-07-09 17:36:41 +02:00
|
|
|
s = Status::Corruption("Table properties missing in " + file_path);
|
|
|
|
ROCKS_LOG_INFO(options_.info_log, "%s", s.ToString().c_str());
|
|
|
|
return s;
|
2020-06-25 04:30:15 +02:00
|
|
|
}
|
|
|
|
}
|
2020-08-12 22:28:43 +02:00
|
|
|
Status BackupEngineImpl::GetFileChecksumsFromManifestInBackup(
|
|
|
|
Env* src_env, const BackupID& backup_id, const BackupMeta* backup,
|
|
|
|
FileChecksumList* checksum_list) {
|
|
|
|
if (checksum_list == nullptr) {
|
|
|
|
return Status::InvalidArgument("checksum_list is nullptr");
|
|
|
|
}
|
|
|
|
|
|
|
|
checksum_list->reset();
|
|
|
|
Status s;
|
|
|
|
|
|
|
|
// Read CURRENT file to get the latest DB MANIFEST filename in backup_id
|
|
|
|
// and then read the the MANIFEST file to obtain the checksum info stored
|
|
|
|
// in the file.
|
|
|
|
std::string current_rel_path =
|
|
|
|
GetPrivateFileRel(backup_id, false /* tmp */, "CURRENT");
|
|
|
|
s = VerifyFileWithCrc32c(src_env, backup, current_rel_path);
|
|
|
|
if (!s.ok()) {
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
|
|
|
std::string manifest_filename;
|
|
|
|
s = ReadFileToString(src_env, GetAbsolutePath(current_rel_path),
|
|
|
|
&manifest_filename);
|
|
|
|
if (!s.ok()) {
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
// Remove tailing '\n' if any
|
|
|
|
while (!manifest_filename.empty() && manifest_filename.back() == '\n') {
|
|
|
|
manifest_filename.pop_back();
|
|
|
|
}
|
|
|
|
|
|
|
|
std::string manifest_rel_path =
|
|
|
|
GetPrivateFileRel(backup_id, false /* tmp */, manifest_filename);
|
|
|
|
s = VerifyFileWithCrc32c(src_env, backup, manifest_rel_path);
|
|
|
|
if (!s.ok()) {
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
|
|
|
s = GetFileChecksumsFromManifest(src_env, GetAbsolutePath(manifest_rel_path),
|
|
|
|
checksum_list);
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
|
|
|
Status BackupEngineImpl::GetFileChecksumsFromManifest(
|
|
|
|
Env* src_env, const std::string& abs_path,
|
|
|
|
FileChecksumList* checksum_list) {
|
|
|
|
if (checksum_list == nullptr) {
|
|
|
|
return Status::InvalidArgument("checksum_list is nullptr");
|
|
|
|
}
|
|
|
|
|
|
|
|
checksum_list->reset();
|
|
|
|
Status s;
|
|
|
|
|
|
|
|
std::unique_ptr<SequentialFileReader> file_reader;
|
|
|
|
{
|
|
|
|
std::unique_ptr<FSSequentialFile> file;
|
|
|
|
const std::shared_ptr<FileSystem>& fs = src_env->GetFileSystem();
|
|
|
|
s = fs->NewSequentialFile(abs_path,
|
|
|
|
fs->OptimizeForManifestRead(FileOptions()), &file,
|
|
|
|
nullptr /* dbg */);
|
|
|
|
if (!s.ok()) {
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
file_reader.reset(new SequentialFileReader(std::move(file), abs_path));
|
|
|
|
}
|
|
|
|
|
|
|
|
LogReporter reporter;
|
|
|
|
reporter.status = &s;
|
|
|
|
log::Reader reader(nullptr, std::move(file_reader), &reporter,
|
|
|
|
true /* checksum */, 0 /* log_number */);
|
|
|
|
Slice record;
|
|
|
|
std::string scratch;
|
|
|
|
// Set of column families initialized with default CF
|
|
|
|
std::unordered_set<uint32_t> cf_set = {0};
|
|
|
|
while (reader.ReadRecord(&record, &scratch) && s.ok()) {
|
|
|
|
VersionEdit edit;
|
|
|
|
s = edit.DecodeFrom(record);
|
|
|
|
if (!s.ok()) {
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
// Check current CF status
|
|
|
|
uint32_t column_family = edit.GetColumnFamily();
|
|
|
|
auto cf_set_itr = cf_set.find(column_family);
|
|
|
|
bool cf_exist = (cf_set_itr != cf_set.end());
|
|
|
|
if (edit.IsColumnFamilyAdd()) {
|
|
|
|
if (cf_exist) {
|
|
|
|
s = Status::Corruption("Manifest adding the same column family twice");
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
cf_set.insert(column_family);
|
|
|
|
} else if (edit.IsColumnFamilyDrop()) {
|
|
|
|
if (!cf_exist) {
|
|
|
|
s = Status::Corruption(
|
|
|
|
"Manifest dropping non-existing column family: " +
|
|
|
|
ToString(column_family));
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
cf_set.erase(cf_set_itr);
|
|
|
|
} else {
|
|
|
|
if (!cf_exist) {
|
|
|
|
s = Status::Corruption("Manifest referencing unknown column family: " +
|
|
|
|
ToString(column_family));
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
assert(cf_set.find(column_family) != cf_set.end());
|
|
|
|
|
|
|
|
// Remove the deleted files from the checksum_list
|
|
|
|
for (const auto& deleted_file : edit.GetDeletedFiles()) {
|
|
|
|
checksum_list->RemoveOneFileChecksum(deleted_file.second);
|
|
|
|
}
|
|
|
|
|
|
|
|
// Add the new files to the checksum_list
|
|
|
|
for (const auto& new_file : edit.GetNewFiles()) {
|
|
|
|
checksum_list->InsertOneFileChecksum(
|
|
|
|
new_file.second.fd.GetNumber(), new_file.second.file_checksum,
|
|
|
|
new_file.second.file_checksum_func_name);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
|
|
|
Status BackupEngineImpl::VerifyFileWithCrc32c(Env* src_env,
|
|
|
|
const BackupMeta* backup,
|
|
|
|
const std::string& rel_path) {
|
|
|
|
const std::shared_ptr<FileInfo> file_info = backup->GetFile(rel_path);
|
|
|
|
if (file_info == nullptr) {
|
|
|
|
return Status::Corruption(rel_path + " is missing");
|
|
|
|
}
|
|
|
|
|
|
|
|
std::string abs_path = GetAbsolutePath(rel_path);
|
|
|
|
std::string expected_checksum = file_info->checksum_hex;
|
|
|
|
std::string actual_checksum;
|
|
|
|
Status s = CalculateChecksum(abs_path, src_env, EnvOptions(),
|
|
|
|
0 /* size_limit */, &actual_checksum);
|
|
|
|
if (!s.ok()) {
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
if (actual_checksum != expected_checksum) {
|
|
|
|
std::string checksum_info("Expected checksum is " + expected_checksum +
|
|
|
|
" while computed checksum is " + actual_checksum);
|
|
|
|
return Status::Corruption("crc32c mismatch for " + rel_path + ": " +
|
|
|
|
checksum_info);
|
|
|
|
}
|
|
|
|
return s;
|
|
|
|
}
|
2020-06-25 04:30:15 +02:00
|
|
|
|
2014-03-17 23:39:23 +01:00
|
|
|
void BackupEngineImpl::DeleteChildren(const std::string& dir,
|
|
|
|
uint32_t file_type_filter) {
|
|
|
|
std::vector<std::string> children;
|
|
|
|
db_env_->GetChildren(dir, &children); // ignore errors
|
|
|
|
|
|
|
|
for (const auto& f : children) {
|
|
|
|
uint64_t number;
|
|
|
|
FileType type;
|
|
|
|
bool ok = ParseFileName(f, &number, &type);
|
|
|
|
if (ok && (file_type_filter & (1 << type))) {
|
|
|
|
// don't delete this file
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
db_env_->DeleteFile(dir + "/" + f); // ignore errors
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2016-03-02 04:33:33 +01:00
|
|
|
Status BackupEngineImpl::InsertPathnameToSizeBytes(
|
2016-04-26 21:33:30 +02:00
|
|
|
const std::string& dir, Env* env,
|
|
|
|
std::unordered_map<std::string, uint64_t>* result) {
|
2016-03-02 04:33:33 +01:00
|
|
|
assert(result != nullptr);
|
|
|
|
std::vector<Env::FileAttributes> files_attrs;
|
2017-02-23 19:46:54 +01:00
|
|
|
Status status = env->FileExists(dir);
|
|
|
|
if (status.ok()) {
|
|
|
|
status = env->GetChildrenFileAttributes(dir, &files_attrs);
|
|
|
|
} else if (status.IsNotFound()) {
|
|
|
|
// Insert no entries can be considered success
|
|
|
|
status = Status::OK();
|
2016-03-02 04:33:33 +01:00
|
|
|
}
|
|
|
|
const bool slash_needed = dir.empty() || dir.back() != '/';
|
|
|
|
for (const auto& file_attrs : files_attrs) {
|
|
|
|
result->emplace(dir + (slash_needed ? "/" : "") + file_attrs.name,
|
|
|
|
file_attrs.size_bytes);
|
|
|
|
}
|
2017-02-23 19:46:54 +01:00
|
|
|
return status;
|
2016-03-02 04:33:33 +01:00
|
|
|
}
|
|
|
|
|
2014-11-13 18:51:41 +01:00
|
|
|
Status BackupEngineImpl::GarbageCollect() {
|
2014-04-25 21:49:29 +02:00
|
|
|
assert(!read_only_);
|
Auto-GarbageCollect on PurgeOldBackups and DeleteBackup (#6015)
Summary:
Only if there is a crash, power failure, or I/O error in
DeleteBackup, shared or private files from the backup might be left
behind that are not cleaned up by PurgeOldBackups or DeleteBackup-- only
by GarbageCollect. This makes the BackupEngine API "leaky by default."
Even if it means a modest performance hit, I think we should make
Delete and Purge do as they say, with ongoing best effort: i.e. future
calls will attempt to finish any incomplete work from earlier calls.
This change does that by having DeleteBackup and PurgeOldBackups do a
GarbageCollect, unless (to minimize performance hit) this BackupEngine
has already done a GarbageCollect and there have been no
deletion-related I/O errors in that GarbageCollect or since then.
Rejected alternative 1: remove meta file last instead of first. This would in theory turn partially deleted backups into corrupted backups, but code changes would be needed to allow the missing files and consider it acceptably corrupt, rather than failing to open the BackupEngine. This might be a reasonable choice, but I mostly rejected it because it doesn't solve the legacy problem of cleaning up existing lingering files.
Rejected alternative 2: use a deletion marker file. If deletion started with creating a file that marks a backup as flagged for deletion, then we could reliably detect partially deleted backups and efficiently finish removing them. In addition to not solving the legacy problem, this could be precarious if there's a disk full situation, and we try to create a new file in order to delete some files. Ugh.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6015
Test Plan: Updated unit tests
Differential Revision: D18401333
Pulled By: pdillinger
fbshipit-source-id: 12944e372ce6809f3f5a4c416c3b321a8927d925
2019-11-09 04:13:41 +01:00
|
|
|
|
|
|
|
// We will make a best effort to remove all garbage even in the presence
|
|
|
|
// of inconsistencies or I/O failures that inhibit finding garbage.
|
|
|
|
Status overall_status = Status::OK();
|
|
|
|
// If all goes well, we don't need another auto-GC this session
|
|
|
|
might_need_garbage_collect_ = false;
|
|
|
|
|
2017-03-16 03:22:52 +01:00
|
|
|
ROCKS_LOG_INFO(options_.info_log, "Starting garbage collection");
|
2019-11-25 23:18:10 +01:00
|
|
|
|
|
|
|
// delete obsolete shared files
|
|
|
|
for (bool with_checksum : {false, true}) {
|
|
|
|
std::vector<std::string> shared_children;
|
|
|
|
{
|
|
|
|
std::string shared_path;
|
|
|
|
if (with_checksum) {
|
|
|
|
shared_path = GetAbsolutePath(GetSharedFileWithChecksumRel());
|
|
|
|
} else {
|
|
|
|
shared_path = GetAbsolutePath(GetSharedFileRel());
|
|
|
|
}
|
|
|
|
auto s = backup_env_->FileExists(shared_path);
|
|
|
|
if (s.ok()) {
|
|
|
|
s = backup_env_->GetChildren(shared_path, &shared_children);
|
|
|
|
} else if (s.IsNotFound()) {
|
|
|
|
s = Status::OK();
|
|
|
|
}
|
|
|
|
if (!s.ok()) {
|
|
|
|
overall_status = s;
|
|
|
|
// Trying again later might work
|
|
|
|
might_need_garbage_collect_ = true;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
for (auto& child : shared_children) {
|
|
|
|
if (child == "." || child == "..") {
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
std::string rel_fname;
|
|
|
|
if (with_checksum) {
|
|
|
|
rel_fname = GetSharedFileWithChecksumRel(child);
|
|
|
|
} else {
|
|
|
|
rel_fname = GetSharedFileRel(child);
|
|
|
|
}
|
|
|
|
auto child_itr = backuped_file_infos_.find(rel_fname);
|
|
|
|
// if it's not refcounted, delete it
|
|
|
|
if (child_itr == backuped_file_infos_.end() ||
|
|
|
|
child_itr->second->refs == 0) {
|
|
|
|
// this might be a directory, but DeleteFile will just fail in that
|
|
|
|
// case, so we're good
|
|
|
|
Status s = backup_env_->DeleteFile(GetAbsolutePath(rel_fname));
|
|
|
|
ROCKS_LOG_INFO(options_.info_log, "Deleting %s -- %s",
|
|
|
|
rel_fname.c_str(), s.ToString().c_str());
|
|
|
|
backuped_file_infos_.erase(rel_fname);
|
Auto-GarbageCollect on PurgeOldBackups and DeleteBackup (#6015)
Summary:
Only if there is a crash, power failure, or I/O error in
DeleteBackup, shared or private files from the backup might be left
behind that are not cleaned up by PurgeOldBackups or DeleteBackup-- only
by GarbageCollect. This makes the BackupEngine API "leaky by default."
Even if it means a modest performance hit, I think we should make
Delete and Purge do as they say, with ongoing best effort: i.e. future
calls will attempt to finish any incomplete work from earlier calls.
This change does that by having DeleteBackup and PurgeOldBackups do a
GarbageCollect, unless (to minimize performance hit) this BackupEngine
has already done a GarbageCollect and there have been no
deletion-related I/O errors in that GarbageCollect or since then.
Rejected alternative 1: remove meta file last instead of first. This would in theory turn partially deleted backups into corrupted backups, but code changes would be needed to allow the missing files and consider it acceptably corrupt, rather than failing to open the BackupEngine. This might be a reasonable choice, but I mostly rejected it because it doesn't solve the legacy problem of cleaning up existing lingering files.
Rejected alternative 2: use a deletion marker file. If deletion started with creating a file that marks a backup as flagged for deletion, then we could reliably detect partially deleted backups and efficiently finish removing them. In addition to not solving the legacy problem, this could be precarious if there's a disk full situation, and we try to create a new file in order to delete some files. Ugh.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6015
Test Plan: Updated unit tests
Differential Revision: D18401333
Pulled By: pdillinger
fbshipit-source-id: 12944e372ce6809f3f5a4c416c3b321a8927d925
2019-11-09 04:13:41 +01:00
|
|
|
if (!s.ok()) {
|
|
|
|
// Trying again later might work
|
|
|
|
might_need_garbage_collect_ = true;
|
|
|
|
}
|
2016-01-06 22:05:24 +01:00
|
|
|
}
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
}
|
2014-11-13 18:51:41 +01:00
|
|
|
}
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
|
2014-11-13 18:51:41 +01:00
|
|
|
// delete obsolete private files
|
|
|
|
std::vector<std::string> private_children;
|
2015-07-14 20:51:36 +02:00
|
|
|
{
|
|
|
|
auto s = backup_env_->GetChildren(GetAbsolutePath(GetPrivateDirRel()),
|
|
|
|
&private_children);
|
|
|
|
if (!s.ok()) {
|
Auto-GarbageCollect on PurgeOldBackups and DeleteBackup (#6015)
Summary:
Only if there is a crash, power failure, or I/O error in
DeleteBackup, shared or private files from the backup might be left
behind that are not cleaned up by PurgeOldBackups or DeleteBackup-- only
by GarbageCollect. This makes the BackupEngine API "leaky by default."
Even if it means a modest performance hit, I think we should make
Delete and Purge do as they say, with ongoing best effort: i.e. future
calls will attempt to finish any incomplete work from earlier calls.
This change does that by having DeleteBackup and PurgeOldBackups do a
GarbageCollect, unless (to minimize performance hit) this BackupEngine
has already done a GarbageCollect and there have been no
deletion-related I/O errors in that GarbageCollect or since then.
Rejected alternative 1: remove meta file last instead of first. This would in theory turn partially deleted backups into corrupted backups, but code changes would be needed to allow the missing files and consider it acceptably corrupt, rather than failing to open the BackupEngine. This might be a reasonable choice, but I mostly rejected it because it doesn't solve the legacy problem of cleaning up existing lingering files.
Rejected alternative 2: use a deletion marker file. If deletion started with creating a file that marks a backup as flagged for deletion, then we could reliably detect partially deleted backups and efficiently finish removing them. In addition to not solving the legacy problem, this could be precarious if there's a disk full situation, and we try to create a new file in order to delete some files. Ugh.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6015
Test Plan: Updated unit tests
Differential Revision: D18401333
Pulled By: pdillinger
fbshipit-source-id: 12944e372ce6809f3f5a4c416c3b321a8927d925
2019-11-09 04:13:41 +01:00
|
|
|
overall_status = s;
|
|
|
|
// Trying again later might work
|
|
|
|
might_need_garbage_collect_ = true;
|
2015-07-14 20:51:36 +02:00
|
|
|
}
|
|
|
|
}
|
2014-11-13 18:51:41 +01:00
|
|
|
for (auto& child : private_children) {
|
2019-11-14 15:18:23 +01:00
|
|
|
if (child == "." || child == "..") {
|
|
|
|
continue;
|
|
|
|
}
|
2019-11-25 23:18:10 +01:00
|
|
|
|
2014-11-13 18:51:41 +01:00
|
|
|
BackupID backup_id = 0;
|
|
|
|
bool tmp_dir = child.find(".tmp") != std::string::npos;
|
|
|
|
sscanf(child.c_str(), "%u", &backup_id);
|
|
|
|
if (!tmp_dir && // if it's tmp_dir, delete it
|
|
|
|
(backup_id == 0 || backups_.find(backup_id) != backups_.end())) {
|
|
|
|
// it's either not a number or it's still alive. continue
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
// here we have to delete the dir and all its children
|
|
|
|
std::string full_private_path =
|
2018-04-21 02:23:34 +02:00
|
|
|
GetAbsolutePath(GetPrivateFileRel(backup_id));
|
2014-11-13 18:51:41 +01:00
|
|
|
std::vector<std::string> subchildren;
|
|
|
|
backup_env_->GetChildren(full_private_path, &subchildren);
|
|
|
|
for (auto& subchild : subchildren) {
|
2019-11-14 15:18:23 +01:00
|
|
|
if (subchild == "." || subchild == "..") {
|
|
|
|
continue;
|
|
|
|
}
|
2014-11-13 18:51:41 +01:00
|
|
|
Status s = backup_env_->DeleteFile(full_private_path + subchild);
|
2017-03-16 03:22:52 +01:00
|
|
|
ROCKS_LOG_INFO(options_.info_log, "Deleting %s -- %s",
|
|
|
|
(full_private_path + subchild).c_str(),
|
|
|
|
s.ToString().c_str());
|
Auto-GarbageCollect on PurgeOldBackups and DeleteBackup (#6015)
Summary:
Only if there is a crash, power failure, or I/O error in
DeleteBackup, shared or private files from the backup might be left
behind that are not cleaned up by PurgeOldBackups or DeleteBackup-- only
by GarbageCollect. This makes the BackupEngine API "leaky by default."
Even if it means a modest performance hit, I think we should make
Delete and Purge do as they say, with ongoing best effort: i.e. future
calls will attempt to finish any incomplete work from earlier calls.
This change does that by having DeleteBackup and PurgeOldBackups do a
GarbageCollect, unless (to minimize performance hit) this BackupEngine
has already done a GarbageCollect and there have been no
deletion-related I/O errors in that GarbageCollect or since then.
Rejected alternative 1: remove meta file last instead of first. This would in theory turn partially deleted backups into corrupted backups, but code changes would be needed to allow the missing files and consider it acceptably corrupt, rather than failing to open the BackupEngine. This might be a reasonable choice, but I mostly rejected it because it doesn't solve the legacy problem of cleaning up existing lingering files.
Rejected alternative 2: use a deletion marker file. If deletion started with creating a file that marks a backup as flagged for deletion, then we could reliably detect partially deleted backups and efficiently finish removing them. In addition to not solving the legacy problem, this could be precarious if there's a disk full situation, and we try to create a new file in order to delete some files. Ugh.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6015
Test Plan: Updated unit tests
Differential Revision: D18401333
Pulled By: pdillinger
fbshipit-source-id: 12944e372ce6809f3f5a4c416c3b321a8927d925
2019-11-09 04:13:41 +01:00
|
|
|
if (!s.ok()) {
|
|
|
|
// Trying again later might work
|
|
|
|
might_need_garbage_collect_ = true;
|
|
|
|
}
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
}
|
2014-11-13 18:51:41 +01:00
|
|
|
// finally delete the private dir
|
|
|
|
Status s = backup_env_->DeleteDir(full_private_path);
|
2017-03-16 03:22:52 +01:00
|
|
|
ROCKS_LOG_INFO(options_.info_log, "Deleting dir %s -- %s",
|
|
|
|
full_private_path.c_str(), s.ToString().c_str());
|
Auto-GarbageCollect on PurgeOldBackups and DeleteBackup (#6015)
Summary:
Only if there is a crash, power failure, or I/O error in
DeleteBackup, shared or private files from the backup might be left
behind that are not cleaned up by PurgeOldBackups or DeleteBackup-- only
by GarbageCollect. This makes the BackupEngine API "leaky by default."
Even if it means a modest performance hit, I think we should make
Delete and Purge do as they say, with ongoing best effort: i.e. future
calls will attempt to finish any incomplete work from earlier calls.
This change does that by having DeleteBackup and PurgeOldBackups do a
GarbageCollect, unless (to minimize performance hit) this BackupEngine
has already done a GarbageCollect and there have been no
deletion-related I/O errors in that GarbageCollect or since then.
Rejected alternative 1: remove meta file last instead of first. This would in theory turn partially deleted backups into corrupted backups, but code changes would be needed to allow the missing files and consider it acceptably corrupt, rather than failing to open the BackupEngine. This might be a reasonable choice, but I mostly rejected it because it doesn't solve the legacy problem of cleaning up existing lingering files.
Rejected alternative 2: use a deletion marker file. If deletion started with creating a file that marks a backup as flagged for deletion, then we could reliably detect partially deleted backups and efficiently finish removing them. In addition to not solving the legacy problem, this could be precarious if there's a disk full situation, and we try to create a new file in order to delete some files. Ugh.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6015
Test Plan: Updated unit tests
Differential Revision: D18401333
Pulled By: pdillinger
fbshipit-source-id: 12944e372ce6809f3f5a4c416c3b321a8927d925
2019-11-09 04:13:41 +01:00
|
|
|
if (!s.ok()) {
|
|
|
|
// Trying again later might work
|
|
|
|
might_need_garbage_collect_ = true;
|
|
|
|
}
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
}
|
2014-11-13 18:51:41 +01:00
|
|
|
|
Auto-GarbageCollect on PurgeOldBackups and DeleteBackup (#6015)
Summary:
Only if there is a crash, power failure, or I/O error in
DeleteBackup, shared or private files from the backup might be left
behind that are not cleaned up by PurgeOldBackups or DeleteBackup-- only
by GarbageCollect. This makes the BackupEngine API "leaky by default."
Even if it means a modest performance hit, I think we should make
Delete and Purge do as they say, with ongoing best effort: i.e. future
calls will attempt to finish any incomplete work from earlier calls.
This change does that by having DeleteBackup and PurgeOldBackups do a
GarbageCollect, unless (to minimize performance hit) this BackupEngine
has already done a GarbageCollect and there have been no
deletion-related I/O errors in that GarbageCollect or since then.
Rejected alternative 1: remove meta file last instead of first. This would in theory turn partially deleted backups into corrupted backups, but code changes would be needed to allow the missing files and consider it acceptably corrupt, rather than failing to open the BackupEngine. This might be a reasonable choice, but I mostly rejected it because it doesn't solve the legacy problem of cleaning up existing lingering files.
Rejected alternative 2: use a deletion marker file. If deletion started with creating a file that marks a backup as flagged for deletion, then we could reliably detect partially deleted backups and efficiently finish removing them. In addition to not solving the legacy problem, this could be precarious if there's a disk full situation, and we try to create a new file in order to delete some files. Ugh.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6015
Test Plan: Updated unit tests
Differential Revision: D18401333
Pulled By: pdillinger
fbshipit-source-id: 12944e372ce6809f3f5a4c416c3b321a8927d925
2019-11-09 04:13:41 +01:00
|
|
|
assert(overall_status.ok() || might_need_garbage_collect_);
|
|
|
|
return overall_status;
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
// ------- BackupMeta class --------
|
|
|
|
|
2015-01-05 14:35:09 +01:00
|
|
|
Status BackupEngineImpl::BackupMeta::AddFile(
|
|
|
|
std::shared_ptr<FileInfo> file_info) {
|
|
|
|
auto itr = file_infos_->find(file_info->filename);
|
2014-01-28 18:43:36 +01:00
|
|
|
if (itr == file_infos_->end()) {
|
2015-01-29 22:43:09 +01:00
|
|
|
auto ret = file_infos_->insert({file_info->filename, file_info});
|
2014-01-28 18:43:36 +01:00
|
|
|
if (ret.second) {
|
2015-01-05 14:35:09 +01:00
|
|
|
itr = ret.first;
|
|
|
|
itr->second->refs = 1;
|
2014-01-28 18:43:36 +01:00
|
|
|
} else {
|
|
|
|
// if this happens, something is seriously wrong
|
|
|
|
return Status::Corruption("In memory metadata insertion error");
|
|
|
|
}
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
} else {
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
if (itr->second->checksum_hex != file_info->checksum_hex) {
|
Fix BackupEngine
Summary:
In D28521 we removed GarbageCollect() from BackupEngine's constructor. The reason was that opening BackupEngine on HDFS was very slow and in most cases we didn't have any garbage. We allowed the user to call GarbageCollect() when it detects some garbage files in his backup directory.
Unfortunately, this left us vulnerable to an interesting issue. Let's say we started a backup and copied files {1, 3} but the backup failed. On another host, we restore DB from backup and generate {1, 3, 5}. Since {1, 3} is already there, we will not overwrite. However, these files might be from a different database so their contents might be different. See internal task t6781803 for more info.
Now, when we're copying files and we discover a file already there, we check:
1. if the file is not referenced from any backups, we overwrite the file.
2. if the file is referenced from other backups AND the checksums don't match, we fail the backup. This will only happen if user is using a single backup directory for backing up two different databases.
3. if the file is referenced from other backups AND the checksums match, it's all good. We skip the copy and go copy the next file.
Test Plan: Added new test to backupable_db_test. The test fails before this patch.
Reviewers: sdong, rven, yhchiang
Reviewed By: yhchiang
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D37599
2015-05-08 02:39:19 +02:00
|
|
|
return Status::Corruption(
|
|
|
|
"Checksum mismatch for existing backup file. Delete old backups and "
|
|
|
|
"try again.");
|
2020-08-12 22:28:43 +02:00
|
|
|
} else if (IsSameChecksumFunc(itr->second->checksum_func_name,
|
|
|
|
file_info->checksum_func_name) &&
|
2020-08-14 22:31:56 +02:00
|
|
|
!itr->second->custom_checksum_hex.empty() &&
|
2020-08-12 22:28:43 +02:00
|
|
|
itr->second->custom_checksum_hex !=
|
|
|
|
file_info->custom_checksum_hex) {
|
|
|
|
return Status::Corruption(
|
|
|
|
"Custom checksum mismatch for existing backup file. Delete old "
|
|
|
|
"backups and try again.");
|
2014-01-28 18:43:36 +01:00
|
|
|
}
|
2015-01-05 14:35:09 +01:00
|
|
|
++itr->second->refs; // increase refcount if already present
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
}
|
2014-01-28 18:43:36 +01:00
|
|
|
|
2015-01-05 14:35:09 +01:00
|
|
|
size_ += file_info->size;
|
|
|
|
files_.push_back(itr->second);
|
|
|
|
|
2014-01-28 18:43:36 +01:00
|
|
|
return Status::OK();
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
}
|
|
|
|
|
2015-07-14 20:51:36 +02:00
|
|
|
Status BackupEngineImpl::BackupMeta::Delete(bool delete_meta) {
|
|
|
|
Status s;
|
2014-01-28 18:43:36 +01:00
|
|
|
for (const auto& file : files_) {
|
2015-01-05 14:35:09 +01:00
|
|
|
--file->refs; // decrease refcount
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
}
|
|
|
|
files_.clear();
|
|
|
|
// delete meta file
|
2015-07-21 02:20:40 +02:00
|
|
|
if (delete_meta) {
|
|
|
|
s = env_->FileExists(meta_filename_);
|
|
|
|
if (s.ok()) {
|
|
|
|
s = env_->DeleteFile(meta_filename_);
|
|
|
|
} else if (s.IsNotFound()) {
|
|
|
|
s = Status::OK(); // nothing to delete
|
|
|
|
}
|
2014-04-25 21:49:29 +02:00
|
|
|
}
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
timestamp_ = 0;
|
2015-07-14 20:51:36 +02:00
|
|
|
return s;
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
}
|
|
|
|
|
2016-04-01 19:56:52 +02:00
|
|
|
Slice kMetaDataPrefix("metadata ");
|
|
|
|
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
// each backup meta file is of the format:
|
|
|
|
// <timestamp>
|
2013-12-11 05:49:28 +01:00
|
|
|
// <seq number>
|
2016-04-01 19:56:52 +02:00
|
|
|
// <metadata(literal string)> <metadata> (optional)
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
// <number of files>
|
2020-05-21 17:10:43 +02:00
|
|
|
// <file1> <crc32(literal string)> <crc32c_value>
|
|
|
|
// <file2> <crc32(literal string)> <crc32c_value>
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
// ...
|
2016-03-02 04:33:33 +01:00
|
|
|
Status BackupEngineImpl::BackupMeta::LoadFromFile(
|
2016-03-03 22:32:20 +01:00
|
|
|
const std::string& backup_dir,
|
2016-03-02 04:33:33 +01:00
|
|
|
const std::unordered_map<std::string, uint64_t>& abs_path_to_size) {
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
assert(Empty());
|
|
|
|
Status s;
|
2018-11-09 20:17:34 +01:00
|
|
|
std::unique_ptr<SequentialFile> backup_meta_file;
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
s = env_->NewSequentialFile(meta_filename_, &backup_meta_file, EnvOptions());
|
|
|
|
if (!s.ok()) {
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
2018-11-09 20:17:34 +01:00
|
|
|
std::unique_ptr<SequentialFileReader> backup_meta_reader(
|
Introduce a new storage specific Env API (#5761)
Summary:
The current Env API encompasses both storage/file operations, as well as OS related operations. Most of the APIs return a Status, which does not have enough metadata about an error, such as whether its retry-able or not, scope (i.e fault domain) of the error etc., that may be required in order to properly handle a storage error. The file APIs also do not provide enough control over the IO SLA, such as timeout, prioritization, hinting about placement and redundancy etc.
This PR separates out the file/storage APIs from Env into a new FileSystem class. The APIs are updated to return an IOStatus with metadata about the error, as well as to take an IOOptions structure as input in order to allow more control over the IO.
The user can set both ```options.env``` and ```options.file_system``` to specify that RocksDB should use the former for OS related operations and the latter for storage operations. Internally, a ```CompositeEnvWrapper``` has been introduced that inherits from ```Env``` and redirects individual methods to either an ```Env``` implementation or the ```FileSystem``` as appropriate. When options are sanitized during ```DB::Open```, ```options.env``` is replaced with a newly allocated ```CompositeEnvWrapper``` instance if both env and file_system have been specified. This way, the rest of the RocksDB code can continue to function as before.
This PR also ports PosixEnv to the new API by splitting it into two - PosixEnv and PosixFileSystem. PosixEnv is defined as a sub-class of CompositeEnvWrapper, and threading/time functions are overridden with Posix specific implementations in order to avoid an extra level of indirection.
The ```CompositeEnvWrapper``` translates ```IOStatus``` return code to ```Status```, and sets the severity to ```kSoftError``` if the io_status is retryable. The error handling code in RocksDB can then recover the DB automatically.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5761
Differential Revision: D18868376
Pulled By: anand1976
fbshipit-source-id: 39efe18a162ea746fabac6360ff529baba48486f
2019-12-13 23:47:08 +01:00
|
|
|
new SequentialFileReader(NewLegacySequentialFileWrapper(backup_meta_file),
|
|
|
|
meta_filename_));
|
2018-11-09 20:17:34 +01:00
|
|
|
std::unique_ptr<char[]> buf(new char[max_backup_meta_file_size_ + 1]);
|
2013-12-10 19:35:06 +01:00
|
|
|
Slice data;
|
Move rate_limiter, write buffering, most perf context instrumentation and most random kill out of Env
Summary: We want to keep Env a think layer for better portability. Less platform dependent codes should be moved out of Env. In this patch, I create a wrapper of file readers and writers, and put rate limiting, write buffering, as well as most perf context instrumentation and random kill out of Env. It will make it easier to maintain multiple Env in the future.
Test Plan: Run all existing unit tests.
Reviewers: anthony, kradhakrishnan, IslamAbdelRahman, yhchiang, igor
Reviewed By: igor
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D42321
2015-07-18 01:16:11 +02:00
|
|
|
s = backup_meta_reader->Read(max_backup_meta_file_size_, &data, buf.get());
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
|
|
|
|
if (!s.ok() || data.size() == max_backup_meta_file_size_) {
|
2014-02-12 20:42:54 +01:00
|
|
|
return s.ok() ? Status::Corruption("File size too big") : s;
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
}
|
|
|
|
buf[data.size()] = 0;
|
|
|
|
|
|
|
|
uint32_t num_files = 0;
|
2015-01-04 05:02:52 +01:00
|
|
|
char *next;
|
|
|
|
timestamp_ = strtoull(data.data(), &next, 10);
|
|
|
|
data.remove_prefix(next - data.data() + 1); // +1 for '\n'
|
|
|
|
sequence_number_ = strtoull(data.data(), &next, 10);
|
|
|
|
data.remove_prefix(next - data.data() + 1); // +1 for '\n'
|
2016-04-01 19:56:52 +02:00
|
|
|
|
|
|
|
if (data.starts_with(kMetaDataPrefix)) {
|
|
|
|
// app metadata present
|
|
|
|
data.remove_prefix(kMetaDataPrefix.size());
|
|
|
|
Slice hex_encoded_metadata = GetSliceUntil(&data, '\n');
|
|
|
|
bool decode_success = hex_encoded_metadata.DecodeHex(&app_metadata_);
|
|
|
|
if (!decode_success) {
|
|
|
|
return Status::Corruption(
|
|
|
|
"Failed to decode stored hex encoded app metadata");
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2015-01-05 14:20:06 +01:00
|
|
|
num_files = static_cast<uint32_t>(strtoul(data.data(), &next, 10));
|
2015-01-04 05:02:52 +01:00
|
|
|
data.remove_prefix(next - data.data() + 1); // +1 for '\n'
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
|
2015-01-05 14:35:09 +01:00
|
|
|
std::vector<std::shared_ptr<FileInfo>> files;
|
2013-12-11 05:49:28 +01:00
|
|
|
|
2020-05-21 17:10:43 +02:00
|
|
|
// WART: The checksums are crc32c, not original crc32
|
2015-01-04 05:02:52 +01:00
|
|
|
Slice checksum_prefix("crc32 ");
|
|
|
|
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
for (uint32_t i = 0; s.ok() && i < num_files; ++i) {
|
2014-01-28 18:43:36 +01:00
|
|
|
auto line = GetSliceUntil(&data, '\n');
|
2020-06-25 04:30:15 +02:00
|
|
|
// filename is relative, i.e., shared/number.sst,
|
|
|
|
// shared_checksum/number.sst, or private/backup_id/number.sst
|
2014-01-28 18:43:36 +01:00
|
|
|
std::string filename = GetSliceUntil(&line, ' ').ToString();
|
|
|
|
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
uint64_t size;
|
2015-01-05 14:35:09 +01:00
|
|
|
const std::shared_ptr<FileInfo> file_info = GetFile(filename);
|
|
|
|
if (file_info) {
|
2015-01-04 05:06:59 +01:00
|
|
|
size = file_info->size;
|
|
|
|
} else {
|
2016-03-03 22:32:20 +01:00
|
|
|
std::string abs_path = backup_dir + "/" + filename;
|
|
|
|
try {
|
|
|
|
size = abs_path_to_size.at(abs_path);
|
2016-03-04 00:08:24 +01:00
|
|
|
} catch (std::out_of_range&) {
|
2016-12-15 01:29:23 +01:00
|
|
|
return Status::Corruption("Size missing for pathname: " + abs_path);
|
2015-01-04 05:06:59 +01:00
|
|
|
}
|
2014-03-05 02:02:25 +01:00
|
|
|
}
|
2014-01-28 18:43:36 +01:00
|
|
|
|
|
|
|
if (line.empty()) {
|
2015-02-27 23:21:16 +01:00
|
|
|
return Status::Corruption("File checksum is missing for " + filename +
|
|
|
|
" in " + meta_filename_);
|
2014-01-28 18:43:36 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
uint32_t checksum_value = 0;
|
2020-08-12 22:28:43 +02:00
|
|
|
std::string checksum_func_name = kUnknownFileChecksumFuncName;
|
2015-01-04 05:02:52 +01:00
|
|
|
if (line.starts_with(checksum_prefix)) {
|
|
|
|
line.remove_prefix(checksum_prefix.size());
|
2020-08-12 22:28:43 +02:00
|
|
|
checksum_func_name = kDefaultBackupFileChecksumFuncName;
|
|
|
|
checksum_value = static_cast<uint32_t>(strtoul(line.data(), nullptr, 10));
|
2020-02-20 21:07:53 +01:00
|
|
|
if (line != ROCKSDB_NAMESPACE::ToString(checksum_value)) {
|
2020-08-12 22:28:43 +02:00
|
|
|
return Status::Corruption("Invalid crc32c checksum value for " +
|
|
|
|
filename + " in " + meta_filename_);
|
2014-01-28 18:43:36 +01:00
|
|
|
}
|
|
|
|
} else {
|
2015-02-27 23:21:16 +01:00
|
|
|
return Status::Corruption("Unknown checksum type for " + filename +
|
|
|
|
" in " + meta_filename_);
|
2014-01-28 18:43:36 +01:00
|
|
|
}
|
|
|
|
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
files.emplace_back(
|
2020-08-12 22:28:43 +02:00
|
|
|
new FileInfo(filename, size, ChecksumInt32ToHex(checksum_value),
|
|
|
|
"" /* custom_checksum_hex */, checksum_func_name));
|
2013-12-11 05:49:28 +01:00
|
|
|
}
|
|
|
|
|
2014-03-05 02:02:25 +01:00
|
|
|
if (s.ok() && data.size() > 0) {
|
|
|
|
// file has to be read completely. if not, we count it as corruption
|
2015-02-27 23:21:16 +01:00
|
|
|
s = Status::Corruption("Tailing data in backup meta file in " +
|
|
|
|
meta_filename_);
|
2014-03-05 02:02:25 +01:00
|
|
|
}
|
|
|
|
|
2013-12-11 05:49:28 +01:00
|
|
|
if (s.ok()) {
|
2015-01-05 14:35:09 +01:00
|
|
|
files_.reserve(files.size());
|
2014-01-28 18:43:36 +01:00
|
|
|
for (const auto& file_info : files) {
|
|
|
|
s = AddFile(file_info);
|
|
|
|
if (!s.ok()) {
|
|
|
|
break;
|
|
|
|
}
|
2013-12-11 05:49:28 +01:00
|
|
|
}
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
2014-01-28 20:26:07 +01:00
|
|
|
Status BackupEngineImpl::BackupMeta::StoreToFile(bool sync) {
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
Status s;
|
2018-11-09 20:17:34 +01:00
|
|
|
std::unique_ptr<WritableFile> backup_meta_file;
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
EnvOptions env_options;
|
|
|
|
env_options.use_mmap_writes = false;
|
2016-12-22 21:51:29 +01:00
|
|
|
env_options.use_direct_writes = false;
|
2018-02-22 02:33:14 +01:00
|
|
|
s = env_->NewWritableFile(meta_tmp_filename_, &backup_meta_file, env_options);
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
if (!s.ok()) {
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
2018-11-09 20:17:34 +01:00
|
|
|
std::unique_ptr<char[]> buf(new char[max_backup_meta_file_size_]);
|
2016-04-08 22:09:19 +02:00
|
|
|
size_t len = 0, buf_size = max_backup_meta_file_size_;
|
2013-12-10 19:35:06 +01:00
|
|
|
len += snprintf(buf.get(), buf_size, "%" PRId64 "\n", timestamp_);
|
2013-12-11 05:49:28 +01:00
|
|
|
len += snprintf(buf.get() + len, buf_size - len, "%" PRIu64 "\n",
|
|
|
|
sequence_number_);
|
2016-04-01 19:56:52 +02:00
|
|
|
if (!app_metadata_.empty()) {
|
|
|
|
std::string hex_encoded_metadata =
|
|
|
|
Slice(app_metadata_).ToString(/* hex */ true);
|
2018-03-22 22:04:06 +01:00
|
|
|
|
2018-08-06 19:57:22 +02:00
|
|
|
// +1 to accommodate newline character
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
size_t hex_meta_strlen =
|
|
|
|
kMetaDataPrefix.ToString().length() + hex_encoded_metadata.length() + 1;
|
2018-03-22 22:04:06 +01:00
|
|
|
if (hex_meta_strlen >= buf_size) {
|
|
|
|
return Status::Corruption("Buffer too small to fit backup metadata");
|
|
|
|
}
|
|
|
|
else if (len + hex_meta_strlen >= buf_size) {
|
|
|
|
backup_meta_file->Append(Slice(buf.get(), len));
|
|
|
|
buf.reset();
|
2018-11-09 20:17:34 +01:00
|
|
|
std::unique_ptr<char[]> new_reset_buf(
|
|
|
|
new char[max_backup_meta_file_size_]);
|
2018-03-22 22:04:06 +01:00
|
|
|
buf.swap(new_reset_buf);
|
|
|
|
len = 0;
|
|
|
|
}
|
2017-12-15 21:22:13 +01:00
|
|
|
len += snprintf(buf.get() + len, buf_size - len, "%s%s\n",
|
|
|
|
kMetaDataPrefix.ToString().c_str(),
|
|
|
|
hex_encoded_metadata.c_str());
|
2016-04-01 19:56:52 +02:00
|
|
|
}
|
2018-03-22 22:04:06 +01:00
|
|
|
|
|
|
|
char writelen_temp[19];
|
2018-04-10 04:21:46 +02:00
|
|
|
if (len + snprintf(writelen_temp, sizeof(writelen_temp),
|
|
|
|
"%" ROCKSDB_PRIszt "\n", files_.size()) >= buf_size) {
|
2018-03-22 22:04:06 +01:00
|
|
|
backup_meta_file->Append(Slice(buf.get(), len));
|
|
|
|
buf.reset();
|
2018-11-09 20:17:34 +01:00
|
|
|
std::unique_ptr<char[]> new_reset_buf(new char[max_backup_meta_file_size_]);
|
2018-03-22 22:04:06 +01:00
|
|
|
buf.swap(new_reset_buf);
|
|
|
|
len = 0;
|
2017-12-15 21:22:13 +01:00
|
|
|
}
|
2018-03-22 22:04:06 +01:00
|
|
|
{
|
|
|
|
const char *const_write = writelen_temp;
|
|
|
|
len += snprintf(buf.get() + len, buf_size - len, "%s", const_write);
|
|
|
|
}
|
|
|
|
|
2014-01-28 18:43:36 +01:00
|
|
|
for (const auto& file : files_) {
|
2020-05-21 17:10:43 +02:00
|
|
|
// use crc32c for now, switch to something else if needed
|
|
|
|
// WART: The checksums are crc32c, not original crc32
|
2018-03-22 22:04:06 +01:00
|
|
|
|
BackupEngine computes table checksums only once if db session ids are available (#7110)
Summary:
BackupEngine requires computing table checksums twice when backing up table files to the `shared_checksum` directory.
The repeated computation can be avoided by utilizing the db session id stored as a part of the table properties.
Filenames of table files in the `shared_checksum` directory depend on the following conditions:
1. the naming scheme is `kOptionalChecksumAndDbSessionId`,
2. `db_session_id` is not empty,
3. checksum is available in the DB manifest.
If 1,2,3 are satisfied, then the filenames will be of the form `<file_number>_<checksum>_<db_session_id>.sst`.
If 1,2 are satisfied, then the filenames will be of the form `<file_number>_<db_session_id>.sst`.
In all other cases, the filenames are of the form `<file_number>_<checksum>_<size>.sst`.
Additionally, if `kOptionalChecksumAndDbSessionId` is used (and not falling back to `kChecksumAndFileSize`), the `<checksum>` appeared in the filenames is hexadecimally encoded, instead of being plain `uint32_t` value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7110
Test Plan: backupable_db_test and manual tests.
Reviewed By: ajkr
Differential Revision: D22508992
Pulled By: gg814
fbshipit-source-id: 5669f0ea9ad5a097f69f6d87aca4abba15032389
2020-07-21 19:34:30 +02:00
|
|
|
size_t newlen =
|
|
|
|
len + file->filename.length() +
|
|
|
|
snprintf(writelen_temp, sizeof(writelen_temp), " crc32 %u\n",
|
|
|
|
ChecksumHexToInt32(file->checksum_hex));
|
2020-08-12 22:28:43 +02:00
|
|
|
const char* const_write = writelen_temp;
|
2018-03-22 22:04:06 +01:00
|
|
|
if (newlen >= buf_size) {
|
|
|
|
backup_meta_file->Append(Slice(buf.get(), len));
|
|
|
|
buf.reset();
|
2018-11-09 20:17:34 +01:00
|
|
|
std::unique_ptr<char[]> new_reset_buf(
|
|
|
|
new char[max_backup_meta_file_size_]);
|
2018-03-22 22:04:06 +01:00
|
|
|
buf.swap(new_reset_buf);
|
|
|
|
len = 0;
|
|
|
|
}
|
|
|
|
len += snprintf(buf.get() + len, buf_size - len, "%s%s",
|
|
|
|
file->filename.c_str(), const_write);
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
}
|
|
|
|
|
2016-04-08 22:09:19 +02:00
|
|
|
s = backup_meta_file->Append(Slice(buf.get(), len));
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
if (s.ok() && sync) {
|
|
|
|
s = backup_meta_file->Sync();
|
|
|
|
}
|
|
|
|
if (s.ok()) {
|
|
|
|
s = backup_meta_file->Close();
|
|
|
|
}
|
|
|
|
if (s.ok()) {
|
2018-02-22 02:33:14 +01:00
|
|
|
s = env_->RenameFile(meta_tmp_filename_, meta_filename_);
|
[RocksDB] BackupableDB
Summary:
In this diff I present you BackupableDB v1. You can easily use it to backup your DB and it will do incremental snapshots for you.
Let's first describe how you would use BackupableDB. It's inheriting StackableDB interface so you can easily construct it with your DB object -- it will add a method RollTheSnapshot() to the DB object. When you call RollTheSnapshot(), current snapshot of the DB will be stored in the backup dir. To restore, you can just call RestoreDBFromBackup() on a BackupableDB (which is a static method) and it will restore all files from the backup dir. In the next version, it will even support automatic backuping every X minutes.
There are multiple things you can configure:
1. backup_env and db_env can be different, which is awesome because then you can easily backup to HDFS or wherever you feel like.
2. sync - if true, it *guarantees* backup consistency on machine reboot
3. number of snapshots to keep - this will keep last N snapshots around if you want, for some reason, be able to restore from an earlier snapshot. All the backuping is done in incremental fashion - if we already have 00010.sst, we will not copy it again. *IMPORTANT* -- This is based on assumption that 00010.sst never changes - two files named 00010.sst from the same DB will always be exactly the same. Is this true? I always copy manifest, current and log files.
4. You can decide if you want to flush the memtables before you backup, or you're fine with backing up the log files -- either way, you get a complete and consistent view of the database at a time of backup.
5. More things you can find in BackupableDBOptions
Here is the directory structure I use:
backup_dir/CURRENT_SNAPSHOT - just 4 bytes holding the latest snapshot
0, 1, 2, ... - files containing serialized version of each snapshot - containing a list of files
files/*.sst - sst files shared between snapshots - if one snapshot references 00010.sst and another one needs to backup it from the DB, it will just reference the same file
files/ 0/, 1/, 2/, ... - snapshot directories containing private snapshot files - current, manifest and log files
All the files are ref counted and deleted immediatelly when they get out of scope.
Some other stuff in this diff:
1. Added GetEnv() method to the DB. Discussed with @haobo and we agreed that it seems right thing to do.
2. Fixed StackableDB interface. The way it was set up before, I was not able to implement BackupableDB.
Test Plan:
I have a unittest, but please don't look at this yet. I just hacked it up to help me with debugging. I will write a lot of good tests and update the diff.
Also, `make asan_check`
Reviewers: dhruba, haobo, emayanke
Reviewed By: dhruba
CC: leveldb, haobo
Differential Revision: https://reviews.facebook.net/D14295
2013-12-09 23:06:52 +01:00
|
|
|
}
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
2014-04-25 21:49:29 +02:00
|
|
|
// -------- BackupEngineReadOnlyImpl ---------
|
|
|
|
class BackupEngineReadOnlyImpl : public BackupEngineReadOnly {
|
|
|
|
public:
|
2020-03-29 04:05:54 +02:00
|
|
|
BackupEngineReadOnlyImpl(const BackupableDBOptions& options, Env* db_env)
|
|
|
|
: backup_engine_(new BackupEngineImpl(options, db_env, true)) {}
|
2014-04-26 23:21:39 +02:00
|
|
|
|
2019-02-14 22:52:47 +01:00
|
|
|
~BackupEngineReadOnlyImpl() override {}
|
2014-04-25 21:49:29 +02:00
|
|
|
|
2016-07-06 21:54:01 +02:00
|
|
|
// The returned BackupInfos are in chronological order, which means the
|
|
|
|
// latest backup comes last.
|
2019-02-14 22:52:47 +01:00
|
|
|
void GetBackupInfo(std::vector<BackupInfo>* backup_info) override {
|
2014-04-25 21:49:29 +02:00
|
|
|
backup_engine_->GetBackupInfo(backup_info);
|
|
|
|
}
|
|
|
|
|
2019-02-14 22:52:47 +01:00
|
|
|
void GetCorruptedBackups(std::vector<BackupID>* corrupt_backup_ids) override {
|
2014-11-13 18:51:41 +01:00
|
|
|
backup_engine_->GetCorruptedBackups(corrupt_backup_ids);
|
|
|
|
}
|
|
|
|
|
2020-03-29 04:05:54 +02:00
|
|
|
using BackupEngineReadOnly::RestoreDBFromBackup;
|
|
|
|
Status RestoreDBFromBackup(const RestoreOptions& options, BackupID backup_id,
|
|
|
|
const std::string& db_dir,
|
|
|
|
const std::string& wal_dir) override {
|
|
|
|
return backup_engine_->RestoreDBFromBackup(options, backup_id, db_dir,
|
|
|
|
wal_dir);
|
2014-04-25 21:49:29 +02:00
|
|
|
}
|
|
|
|
|
2020-03-29 04:05:54 +02:00
|
|
|
using BackupEngineReadOnly::RestoreDBFromLatestBackup;
|
|
|
|
Status RestoreDBFromLatestBackup(const RestoreOptions& options,
|
|
|
|
const std::string& db_dir,
|
|
|
|
const std::string& wal_dir) override {
|
|
|
|
return backup_engine_->RestoreDBFromLatestBackup(options, db_dir, wal_dir);
|
2014-04-25 21:49:29 +02:00
|
|
|
}
|
|
|
|
|
2020-06-26 20:39:43 +02:00
|
|
|
Status VerifyBackup(BackupID backup_id,
|
2020-07-01 03:45:36 +02:00
|
|
|
bool verify_with_checksum = false) override {
|
2020-06-26 20:39:43 +02:00
|
|
|
return backup_engine_->VerifyBackup(backup_id, verify_with_checksum);
|
2015-09-02 18:03:55 +02:00
|
|
|
}
|
|
|
|
|
2015-07-14 20:51:36 +02:00
|
|
|
Status Initialize() { return backup_engine_->Initialize(); }
|
|
|
|
|
2014-04-25 21:49:29 +02:00
|
|
|
private:
|
2014-04-26 23:21:39 +02:00
|
|
|
std::unique_ptr<BackupEngineImpl> backup_engine_;
|
2014-04-25 21:49:29 +02:00
|
|
|
};
|
|
|
|
|
2020-03-29 04:05:54 +02:00
|
|
|
Status BackupEngineReadOnly::Open(const BackupableDBOptions& options, Env* env,
|
2015-02-27 23:03:56 +01:00
|
|
|
BackupEngineReadOnly** backup_engine_ptr) {
|
|
|
|
if (options.destroy_old_data) {
|
|
|
|
return Status::InvalidArgument(
|
|
|
|
"Can't destroy old data with ReadOnly BackupEngine");
|
|
|
|
}
|
2015-07-14 20:51:36 +02:00
|
|
|
std::unique_ptr<BackupEngineReadOnlyImpl> backup_engine(
|
2020-03-29 04:05:54 +02:00
|
|
|
new BackupEngineReadOnlyImpl(options, env));
|
2015-07-14 20:51:36 +02:00
|
|
|
auto s = backup_engine->Initialize();
|
|
|
|
if (!s.ok()) {
|
|
|
|
*backup_engine_ptr = nullptr;
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
*backup_engine_ptr = backup_engine.release();
|
2015-02-27 23:03:56 +01:00
|
|
|
return Status::OK();
|
|
|
|
}
|
|
|
|
|
2020-02-20 21:07:53 +01:00
|
|
|
} // namespace ROCKSDB_NAMESPACE
|
2014-04-15 22:39:26 +02:00
|
|
|
|
|
|
|
#endif // ROCKSDB_LITE
|