Add experimental API MarkForCompaction()
Summary:
Some Mongo+Rocks datasets in Parse's environment are not doing compactions very frequently. During the quiet period (with no IO), we'd like to schedule compactions so that our reads become faster. Also, aggressively compacting during quiet periods helps when write bursts happen. In addition, we also want to compact files that are containing deleted key ranges (like old oplog keys).
All of this is currently not possible with CompactRange() because it's single-threaded and blocks all other compactions from happening. Running CompactRange() risks an issue of blocking writes because we generate too much Level 0 files before the compaction is over. Stopping writes is very dangerous because they hold transaction locks. We tried running manual compaction once on Mongo+Rocks and everything fell apart.
MarkForCompaction() solves all of those problems. This is very light-weight manual compaction. It is lower priority than automatic compactions, which means it shouldn't interfere with background process keeping the LSM tree clean. However, if no automatic compactions need to be run (or we have extra background threads available), we will start compacting files that are marked for compaction.
Test Plan: added a new unit test
Reviewers: yhchiang, rven, MarkCallaghan, sdong
Reviewed By: sdong
Subscribers: yoshinorim, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D37083
2015-04-18 01:44:45 +02:00
|
|
|
// Copyright (c) 2013, Facebook, Inc. All rights reserved.
|
|
|
|
// This source code is licensed under the BSD-style license found in the
|
|
|
|
// LICENSE file in the root directory of this source tree. An additional grant
|
|
|
|
// of patent rights can be found in the PATENTS file in the same directory.
|
|
|
|
//
|
|
|
|
// 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.
|
|
|
|
|
|
|
|
#include "db/db_impl.h"
|
|
|
|
|
|
|
|
#ifndef __STDC_FORMAT_MACROS
|
|
|
|
#define __STDC_FORMAT_MACROS
|
|
|
|
#endif
|
|
|
|
|
Implement DB::PromoteL0 method
Summary:
This diff implements a new `DB` method `PromoteL0` which moves all files in L0
to a given level skipping compaction, provided that the files have disjoint
ranges and all levels up to the target level are empty.
This method provides finer-grain control for trivial compactions, and it is
useful for bulk-loading pre-sorted keys. Compared to D34797, it does not change
the semantics of an existing operation, which can impact existing code.
PromoteL0 is designed to work well in combination with the proposed
`GetSstFileWriter`/`AddFile` interface, enabling to "design" the level structure
by populating one level at a time. Such fine-grained control can be very useful
for static or mostly-static databases.
Test Plan: `make check`
Reviewers: IslamAbdelRahman, philipp, MarkCallaghan, yhchiang, igor, sdong
Reviewed By: sdong
Subscribers: dhruba
Differential Revision: https://reviews.facebook.net/D37107
2015-04-23 21:10:36 +02:00
|
|
|
#include <inttypes.h>
|
Add experimental API MarkForCompaction()
Summary:
Some Mongo+Rocks datasets in Parse's environment are not doing compactions very frequently. During the quiet period (with no IO), we'd like to schedule compactions so that our reads become faster. Also, aggressively compacting during quiet periods helps when write bursts happen. In addition, we also want to compact files that are containing deleted key ranges (like old oplog keys).
All of this is currently not possible with CompactRange() because it's single-threaded and blocks all other compactions from happening. Running CompactRange() risks an issue of blocking writes because we generate too much Level 0 files before the compaction is over. Stopping writes is very dangerous because they hold transaction locks. We tried running manual compaction once on Mongo+Rocks and everything fell apart.
MarkForCompaction() solves all of those problems. This is very light-weight manual compaction. It is lower priority than automatic compactions, which means it shouldn't interfere with background process keeping the LSM tree clean. However, if no automatic compactions need to be run (or we have extra background threads available), we will start compacting files that are marked for compaction.
Test Plan: added a new unit test
Reviewers: yhchiang, rven, MarkCallaghan, sdong
Reviewed By: sdong
Subscribers: yoshinorim, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D37083
2015-04-18 01:44:45 +02:00
|
|
|
#include <vector>
|
|
|
|
|
|
|
|
#include "db/column_family.h"
|
Implement DB::PromoteL0 method
Summary:
This diff implements a new `DB` method `PromoteL0` which moves all files in L0
to a given level skipping compaction, provided that the files have disjoint
ranges and all levels up to the target level are empty.
This method provides finer-grain control for trivial compactions, and it is
useful for bulk-loading pre-sorted keys. Compared to D34797, it does not change
the semantics of an existing operation, which can impact existing code.
PromoteL0 is designed to work well in combination with the proposed
`GetSstFileWriter`/`AddFile` interface, enabling to "design" the level structure
by populating one level at a time. Such fine-grained control can be very useful
for static or mostly-static databases.
Test Plan: `make check`
Reviewers: IslamAbdelRahman, philipp, MarkCallaghan, yhchiang, igor, sdong
Reviewed By: sdong
Subscribers: dhruba
Differential Revision: https://reviews.facebook.net/D37107
2015-04-23 21:10:36 +02:00
|
|
|
#include "db/job_context.h"
|
Add experimental API MarkForCompaction()
Summary:
Some Mongo+Rocks datasets in Parse's environment are not doing compactions very frequently. During the quiet period (with no IO), we'd like to schedule compactions so that our reads become faster. Also, aggressively compacting during quiet periods helps when write bursts happen. In addition, we also want to compact files that are containing deleted key ranges (like old oplog keys).
All of this is currently not possible with CompactRange() because it's single-threaded and blocks all other compactions from happening. Running CompactRange() risks an issue of blocking writes because we generate too much Level 0 files before the compaction is over. Stopping writes is very dangerous because they hold transaction locks. We tried running manual compaction once on Mongo+Rocks and everything fell apart.
MarkForCompaction() solves all of those problems. This is very light-weight manual compaction. It is lower priority than automatic compactions, which means it shouldn't interfere with background process keeping the LSM tree clean. However, if no automatic compactions need to be run (or we have extra background threads available), we will start compacting files that are marked for compaction.
Test Plan: added a new unit test
Reviewers: yhchiang, rven, MarkCallaghan, sdong
Reviewed By: sdong
Subscribers: yoshinorim, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D37083
2015-04-18 01:44:45 +02:00
|
|
|
#include "db/version_set.h"
|
|
|
|
#include "rocksdb/status.h"
|
|
|
|
|
|
|
|
namespace rocksdb {
|
|
|
|
|
|
|
|
#ifndef ROCKSDB_LITE
|
|
|
|
Status DBImpl::SuggestCompactRange(ColumnFamilyHandle* column_family,
|
|
|
|
const Slice* begin, const Slice* end) {
|
|
|
|
auto cfh = reinterpret_cast<ColumnFamilyHandleImpl*>(column_family);
|
|
|
|
auto cfd = cfh->cfd();
|
|
|
|
InternalKey start_key, end_key;
|
|
|
|
if (begin != nullptr) {
|
2015-04-24 03:08:37 +02:00
|
|
|
start_key.SetMaxPossibleForUserKey(*begin);
|
Add experimental API MarkForCompaction()
Summary:
Some Mongo+Rocks datasets in Parse's environment are not doing compactions very frequently. During the quiet period (with no IO), we'd like to schedule compactions so that our reads become faster. Also, aggressively compacting during quiet periods helps when write bursts happen. In addition, we also want to compact files that are containing deleted key ranges (like old oplog keys).
All of this is currently not possible with CompactRange() because it's single-threaded and blocks all other compactions from happening. Running CompactRange() risks an issue of blocking writes because we generate too much Level 0 files before the compaction is over. Stopping writes is very dangerous because they hold transaction locks. We tried running manual compaction once on Mongo+Rocks and everything fell apart.
MarkForCompaction() solves all of those problems. This is very light-weight manual compaction. It is lower priority than automatic compactions, which means it shouldn't interfere with background process keeping the LSM tree clean. However, if no automatic compactions need to be run (or we have extra background threads available), we will start compacting files that are marked for compaction.
Test Plan: added a new unit test
Reviewers: yhchiang, rven, MarkCallaghan, sdong
Reviewed By: sdong
Subscribers: yoshinorim, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D37083
2015-04-18 01:44:45 +02:00
|
|
|
}
|
|
|
|
if (end != nullptr) {
|
2015-04-24 03:08:37 +02:00
|
|
|
end_key.SetMinPossibleForUserKey(*end);
|
Add experimental API MarkForCompaction()
Summary:
Some Mongo+Rocks datasets in Parse's environment are not doing compactions very frequently. During the quiet period (with no IO), we'd like to schedule compactions so that our reads become faster. Also, aggressively compacting during quiet periods helps when write bursts happen. In addition, we also want to compact files that are containing deleted key ranges (like old oplog keys).
All of this is currently not possible with CompactRange() because it's single-threaded and blocks all other compactions from happening. Running CompactRange() risks an issue of blocking writes because we generate too much Level 0 files before the compaction is over. Stopping writes is very dangerous because they hold transaction locks. We tried running manual compaction once on Mongo+Rocks and everything fell apart.
MarkForCompaction() solves all of those problems. This is very light-weight manual compaction. It is lower priority than automatic compactions, which means it shouldn't interfere with background process keeping the LSM tree clean. However, if no automatic compactions need to be run (or we have extra background threads available), we will start compacting files that are marked for compaction.
Test Plan: added a new unit test
Reviewers: yhchiang, rven, MarkCallaghan, sdong
Reviewed By: sdong
Subscribers: yoshinorim, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D37083
2015-04-18 01:44:45 +02:00
|
|
|
}
|
|
|
|
{
|
|
|
|
InstrumentedMutexLock l(&mutex_);
|
|
|
|
auto vstorage = cfd->current()->storage_info();
|
2015-04-29 22:35:48 +02:00
|
|
|
for (int level = 0; level < vstorage->num_non_empty_levels() - 1; ++level) {
|
Add experimental API MarkForCompaction()
Summary:
Some Mongo+Rocks datasets in Parse's environment are not doing compactions very frequently. During the quiet period (with no IO), we'd like to schedule compactions so that our reads become faster. Also, aggressively compacting during quiet periods helps when write bursts happen. In addition, we also want to compact files that are containing deleted key ranges (like old oplog keys).
All of this is currently not possible with CompactRange() because it's single-threaded and blocks all other compactions from happening. Running CompactRange() risks an issue of blocking writes because we generate too much Level 0 files before the compaction is over. Stopping writes is very dangerous because they hold transaction locks. We tried running manual compaction once on Mongo+Rocks and everything fell apart.
MarkForCompaction() solves all of those problems. This is very light-weight manual compaction. It is lower priority than automatic compactions, which means it shouldn't interfere with background process keeping the LSM tree clean. However, if no automatic compactions need to be run (or we have extra background threads available), we will start compacting files that are marked for compaction.
Test Plan: added a new unit test
Reviewers: yhchiang, rven, MarkCallaghan, sdong
Reviewed By: sdong
Subscribers: yoshinorim, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D37083
2015-04-18 01:44:45 +02:00
|
|
|
std::vector<FileMetaData*> inputs;
|
|
|
|
vstorage->GetOverlappingInputs(
|
|
|
|
level, begin == nullptr ? nullptr : &start_key,
|
|
|
|
end == nullptr ? nullptr : &end_key, &inputs);
|
|
|
|
for (auto f : inputs) {
|
|
|
|
f->marked_for_compaction = true;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
// Since we have some more files to compact, we should also recompute
|
|
|
|
// compaction score
|
|
|
|
vstorage->ComputeCompactionScore(*cfd->GetLatestMutableCFOptions(),
|
|
|
|
CompactionOptionsFIFO());
|
|
|
|
SchedulePendingCompaction(cfd);
|
|
|
|
MaybeScheduleFlushOrCompaction();
|
|
|
|
}
|
|
|
|
return Status::OK();
|
|
|
|
}
|
Implement DB::PromoteL0 method
Summary:
This diff implements a new `DB` method `PromoteL0` which moves all files in L0
to a given level skipping compaction, provided that the files have disjoint
ranges and all levels up to the target level are empty.
This method provides finer-grain control for trivial compactions, and it is
useful for bulk-loading pre-sorted keys. Compared to D34797, it does not change
the semantics of an existing operation, which can impact existing code.
PromoteL0 is designed to work well in combination with the proposed
`GetSstFileWriter`/`AddFile` interface, enabling to "design" the level structure
by populating one level at a time. Such fine-grained control can be very useful
for static or mostly-static databases.
Test Plan: `make check`
Reviewers: IslamAbdelRahman, philipp, MarkCallaghan, yhchiang, igor, sdong
Reviewed By: sdong
Subscribers: dhruba
Differential Revision: https://reviews.facebook.net/D37107
2015-04-23 21:10:36 +02:00
|
|
|
|
|
|
|
Status DBImpl::PromoteL0(ColumnFamilyHandle* column_family, int target_level) {
|
|
|
|
assert(column_family);
|
|
|
|
|
|
|
|
if (target_level < 1) {
|
|
|
|
Log(InfoLogLevel::INFO_LEVEL, db_options_.info_log,
|
|
|
|
"PromoteL0 FAILED. Invalid target level %d\n", target_level);
|
|
|
|
return Status::InvalidArgument("Invalid target level");
|
|
|
|
}
|
|
|
|
|
|
|
|
Status status;
|
|
|
|
VersionEdit edit;
|
|
|
|
JobContext job_context(next_job_id_.fetch_add(1), true);
|
|
|
|
{
|
|
|
|
InstrumentedMutexLock l(&mutex_);
|
|
|
|
auto* cfd = static_cast<ColumnFamilyHandleImpl*>(column_family)->cfd();
|
|
|
|
const auto* vstorage = cfd->current()->storage_info();
|
|
|
|
|
|
|
|
if (target_level >= vstorage->num_levels()) {
|
|
|
|
Log(InfoLogLevel::INFO_LEVEL, db_options_.info_log,
|
|
|
|
"PromoteL0 FAILED. Target level %d does not exist\n", target_level);
|
|
|
|
job_context.Clean();
|
|
|
|
return Status::InvalidArgument("Target level does not exist");
|
|
|
|
}
|
|
|
|
|
|
|
|
// Sort L0 files by range.
|
|
|
|
const InternalKeyComparator* icmp = &cfd->internal_comparator();
|
|
|
|
auto l0_files = vstorage->LevelFiles(0);
|
|
|
|
std::sort(l0_files.begin(), l0_files.end(),
|
|
|
|
[icmp](FileMetaData* f1, FileMetaData* f2) {
|
|
|
|
return icmp->Compare(f1->largest, f2->largest) < 0;
|
|
|
|
});
|
|
|
|
|
|
|
|
// Check that no L0 file is being compacted and that they have
|
|
|
|
// non-overlapping ranges.
|
|
|
|
for (size_t i = 0; i < l0_files.size(); ++i) {
|
|
|
|
auto f = l0_files[i];
|
|
|
|
if (f->being_compacted) {
|
|
|
|
Log(InfoLogLevel::INFO_LEVEL, db_options_.info_log,
|
|
|
|
"PromoteL0 FAILED. File %" PRIu64 " being compacted\n",
|
|
|
|
f->fd.GetNumber());
|
|
|
|
job_context.Clean();
|
|
|
|
return Status::InvalidArgument("PromoteL0 called during L0 compaction");
|
|
|
|
}
|
|
|
|
|
|
|
|
if (i == 0) continue;
|
|
|
|
auto prev_f = l0_files[i - 1];
|
|
|
|
if (icmp->Compare(prev_f->largest, f->smallest) >= 0) {
|
|
|
|
Log(InfoLogLevel::INFO_LEVEL, db_options_.info_log,
|
|
|
|
"PromoteL0 FAILED. Files %" PRIu64 " and %" PRIu64
|
|
|
|
" have overlapping ranges\n",
|
|
|
|
prev_f->fd.GetNumber(), f->fd.GetNumber());
|
|
|
|
job_context.Clean();
|
|
|
|
return Status::InvalidArgument("L0 has overlapping files");
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
// Check that all levels up to target_level are empty.
|
|
|
|
for (int level = 1; level <= target_level; ++level) {
|
|
|
|
if (vstorage->NumLevelFiles(level) > 0) {
|
|
|
|
Log(InfoLogLevel::INFO_LEVEL, db_options_.info_log,
|
|
|
|
"PromoteL0 FAILED. Level %d not empty\n", level);
|
|
|
|
job_context.Clean();
|
|
|
|
return Status::InvalidArgument(
|
|
|
|
"All levels up to target_level "
|
|
|
|
"must be empty");
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
edit.SetColumnFamily(cfd->GetID());
|
|
|
|
for (const auto& f : l0_files) {
|
|
|
|
edit.DeleteFile(0, f->fd.GetNumber());
|
|
|
|
edit.AddFile(target_level, f->fd.GetNumber(), f->fd.GetPathId(),
|
|
|
|
f->fd.GetFileSize(), f->smallest, f->largest,
|
2015-06-04 21:03:40 +02:00
|
|
|
f->smallest_seqno, f->largest_seqno,
|
|
|
|
f->marked_for_compaction);
|
Implement DB::PromoteL0 method
Summary:
This diff implements a new `DB` method `PromoteL0` which moves all files in L0
to a given level skipping compaction, provided that the files have disjoint
ranges and all levels up to the target level are empty.
This method provides finer-grain control for trivial compactions, and it is
useful for bulk-loading pre-sorted keys. Compared to D34797, it does not change
the semantics of an existing operation, which can impact existing code.
PromoteL0 is designed to work well in combination with the proposed
`GetSstFileWriter`/`AddFile` interface, enabling to "design" the level structure
by populating one level at a time. Such fine-grained control can be very useful
for static or mostly-static databases.
Test Plan: `make check`
Reviewers: IslamAbdelRahman, philipp, MarkCallaghan, yhchiang, igor, sdong
Reviewed By: sdong
Subscribers: dhruba
Differential Revision: https://reviews.facebook.net/D37107
2015-04-23 21:10:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
status = versions_->LogAndApply(cfd, *cfd->GetLatestMutableCFOptions(),
|
|
|
|
&edit, &mutex_, directories_.GetDbDir());
|
|
|
|
if (status.ok()) {
|
2015-06-17 21:37:59 +02:00
|
|
|
InstallSuperVersionAndScheduleWorkWrapper(
|
|
|
|
cfd, &job_context, *cfd->GetLatestMutableCFOptions());
|
Implement DB::PromoteL0 method
Summary:
This diff implements a new `DB` method `PromoteL0` which moves all files in L0
to a given level skipping compaction, provided that the files have disjoint
ranges and all levels up to the target level are empty.
This method provides finer-grain control for trivial compactions, and it is
useful for bulk-loading pre-sorted keys. Compared to D34797, it does not change
the semantics of an existing operation, which can impact existing code.
PromoteL0 is designed to work well in combination with the proposed
`GetSstFileWriter`/`AddFile` interface, enabling to "design" the level structure
by populating one level at a time. Such fine-grained control can be very useful
for static or mostly-static databases.
Test Plan: `make check`
Reviewers: IslamAbdelRahman, philipp, MarkCallaghan, yhchiang, igor, sdong
Reviewed By: sdong
Subscribers: dhruba
Differential Revision: https://reviews.facebook.net/D37107
2015-04-23 21:10:36 +02:00
|
|
|
}
|
|
|
|
} // lock released here
|
|
|
|
LogFlush(db_options_.info_log);
|
|
|
|
job_context.Clean();
|
|
|
|
|
|
|
|
return status;
|
|
|
|
}
|
Add experimental API MarkForCompaction()
Summary:
Some Mongo+Rocks datasets in Parse's environment are not doing compactions very frequently. During the quiet period (with no IO), we'd like to schedule compactions so that our reads become faster. Also, aggressively compacting during quiet periods helps when write bursts happen. In addition, we also want to compact files that are containing deleted key ranges (like old oplog keys).
All of this is currently not possible with CompactRange() because it's single-threaded and blocks all other compactions from happening. Running CompactRange() risks an issue of blocking writes because we generate too much Level 0 files before the compaction is over. Stopping writes is very dangerous because they hold transaction locks. We tried running manual compaction once on Mongo+Rocks and everything fell apart.
MarkForCompaction() solves all of those problems. This is very light-weight manual compaction. It is lower priority than automatic compactions, which means it shouldn't interfere with background process keeping the LSM tree clean. However, if no automatic compactions need to be run (or we have extra background threads available), we will start compacting files that are marked for compaction.
Test Plan: added a new unit test
Reviewers: yhchiang, rven, MarkCallaghan, sdong
Reviewed By: sdong
Subscribers: yoshinorim, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D37083
2015-04-18 01:44:45 +02:00
|
|
|
#endif // ROCKSDB_LITE
|
|
|
|
|
|
|
|
} // namespace rocksdb
|