Summary:
CPU profiling reveals GetApproximateSizes as a bottleneck for performance. The current implementation is sub-optimal, it scans every file in every level to compute the result.
We can take advantage of the fact that all levels above 0 are sorted in the increasing order of key ranges and use binary search to locate the starting index. This can reduce the number of comparisons required to compute the result.
Test Plan: We have good test coverage. Run the tests.
Reviewers: sdong, igor, rven, dynamike
Subscribers: dynamike, maykov, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D37755
Summary:
Added these events:
* Recovery start, finish and also when recovery creates a file
* Trivial move
* Compaction start, finish and when compaction creates a file
* Flush start, finish
Also includes small fix to EventLogger
Also added option ROCKSDB_PRINT_EVENTS_TO_STDOUT which is useful when we debug things. I've spent far too much time chasing LOG files.
Still didn't get sst table properties in JSON. They are written very deeply into the stack. I'll address in separate diff.
TODO:
* Write specification. Let's first use this for a while and figure out what's good data to put here, too. After that we'll write spec
* Write tools that parse and analyze LOGs. This can be in python or go. Good intern task.
Test Plan: Ran db_bench with ROCKSDB_PRINT_EVENTS_TO_STDOUT. Here's the output: https://phabricator.fb.com/P19811976
Reviewers: sdong, yhchiang, rven, MarkCallaghan, kradhakrishnan, anthony
Reviewed By: anthony
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D37521
Summary:
CompactRange for universal compaction with num_levels > 1 seems to have a bug. The unit test also has a bug so it doesn't capture the problem.
Fix it. Revert the compact range to the logic equivalent to num_levels=1. Always compact all files together.
It should also fix DBTest.IncreaseUniversalCompactionNumLevels. The issue was that options.write_buffer_size = 100 << 10 and options.write_buffer_size = 100 << 10 are not used in later test scenarios. So write_buffer_size of 4MB was used. The compaction trigger condition is not anymore obvious as expected.
Test Plan: Run the new test and all test suites
Reviewers: yhchiang, rven, kradhakrishnan, anthony, igor
Reviewed By: igor
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D37551
Summary:
Based on feedback from D37083.
Are all of these correct? In some spaces it seems like we're doing SetMaxPossibleForUserKey() although we want the smallest possible internal key for user key.
Test Plan: make check
Reviewers: sdong
Reviewed By: sdong
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D37341
Summary: D36669 introduces a bug that trivial moved data is not going to specific level but the next level, which will incorrectly be level 1 for level 0 compaciton if base level is not level 1. Fixing it by appreciating the output level
Test Plan: Run all tests
Reviewers: MarkCallaghan, rven, yhchiang, igor
Reviewed By: igor
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D37119
Summary:
The goal of this diff is to make Compaction class easier to use. This should also make new compaction algorithms easier to write (like CompactFiles from @yhchiang and dynamic leveled and multi-leveled universal from @sdong).
Here are couple of things demonstrating that Compaction class is hard to use:
1. we have two constructors of Compaction class
2. there's this thing called grandparents_, but it appears to only be setup for leveled compaction and not compactfiles
3. it's easy to introduce a subtle and dangerous bug like this: D36225
4. SetupBottomMostLevel() is hard to understand and it shouldn't be. See this comment: afbafeaeae/db/compaction.cc (L236-L241). It also made it harder for @yhchiang to write CompactFiles, as evidenced by this: afbafeaeae/db/compaction_picker.cc (L204-L210)
The problem is that we create Compaction object, which holds a lot of state, and then pass it around to some functions. After those functions are done mutating, then we call couple of functions on Compaction object, like SetupBottommostLevel() and MarkFilesBeingCompacted(). It is very hard to see what's happening with all that Compaction's state while it's travelling across different functions. If you're writing a new PickCompaction() function you need to try really hard to understand what are all the functions you need to run on Compaction object and what state you need to setup.
My proposed solution is to make important parts of Compaction immutable after construction. PickCompaction() should calculate compaction inputs and then pass them onto Compaction object once they are finalized. That makes it easy to create a new compaction -- just provide all the parameters to the constructor and you're done. No need to call confusing functions after you created your object.
This diff doesn't fully achieve that goal, but it comes pretty close. Here are some of the changes:
* have one Compaction constructor instead of two.
* inputs_ is constant after construction
* MarkFilesBeingCompacted() is now private to Compaction class and automatically called on construction/destruction.
* SetupBottommostLevel() is gone. Compaction figures it out on its own based on the input.
* CompactionPicker's functions are not passing around Compaction object anymore. They are only passing around the state that they need.
Test Plan:
make check
make asan_check
make valgrind_check
Reviewers: rven, anthony, sdong, yhchiang
Reviewed By: yhchiang
Subscribers: sdong, yhchiang, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D36687
Summary:
Fix a compile error in ROCKSDB_LITE in db/db_impl.cc
related to internal_stats.
Test Plan: make OPT=-DROCKSDB_LITE shared_lib
Reviewers: sdong, igor
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D36819
Summary: Now EnvOptions uses unsanitized DB options. bytes_per_sync is tuned off when rate_limiter is used, but this change doesn't take effort.
Test Plan: See different I/O pattern in db_bench running fillseq.
Reviewers: yhchiang, kradhakrishnan, rven, anthony, igor
Reviewed By: igor
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D36723
Summary: Now we add warnings when user configures compression and the compression is not supported.
Test Plan:
Configured compression to non-supported values. Observed messages in my log:
2015/03/26-12:17:57.586341 7ffb8a496840 [WARN] Compression type chosen for level 2 is not supported: LZ4. RocksDB will not compress data on level 2.
2015/03/26-12:19:10.768045 7f36f15c5840 [WARN] Compression type chosen is not supported: LZ4. RocksDB will not compress data.
Reviewers: rven, sdong, yhchiang
Reviewed By: sdong
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D35979
Summary:
Currently users have no idea a key is add, delete or merge from TablePropertiesCollector call back. Add a new function to add it.
Also refactor the codes so that
(1) make table property collector and internal table property collector two separate data structures with the later one now exposed
(2) table builders only receive internal table properties
Test Plan: Add cases in table_properties_collector_test to cover both of old and new ways of using TablePropertiesCollector.
Reviewers: yhchiang, igor.sugak, rven, igor
Reviewed By: rven, igor
Subscribers: meyering, yoshinorim, maykov, leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D35373
Summary:
With this change, we use L1 and up to store compaction outputs in universal compaction.
The compaction pick logic stays the same. Outputs are stored in the largest "level" as possible.
If options.num_levels=1, it behaves all the same as now.
Test Plan:
1) convert most of existing unit tests for universal comapaction to include the option of one level and multiple levels.
2) add a unit test to cover parallel compaction in universal compaction and run it in one level and multiple levels
3) add unit test to migrate from multiple level setting back to one level setting
4) add a unit test to insert keys to trigger multiple rounds of compactions and verify results.
Reviewers: rven, kradhakrishnan, yhchiang, igor
Reviewed By: igor
Subscribers: meyering, leveldb, MarkCallaghan, dhruba
Differential Revision: https://reviews.facebook.net/D34539
Summary:
Cleaning up log files can do heavy IO, since we call ftruncate() in the destructor. We don't want to call ftruncate() in user threads.
This diff moves cleaning to background threads (flush and compaction)
Test Plan: make check, will also run valgrind
Reviewers: yhchiang, rven, MarkCallaghan, sdong
Reviewed By: sdong
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D36177
Summary:
Assign the string properties to const string variables under the
DB::Properties namespace. This helps catch typos during compilation and
also consolidates the property definition in one place.
Test Plan: Run rocksdb unit tests
Reviewers: sdong, yoshinorim, igor
Subscribers: dhruba
Differential Revision: https://reviews.facebook.net/D35991
Summary: It's useful to know if we have compression support or no
Test Plan:
Observed this in my LOG:
2015/03/26-10:34:35.460681 7f5b322b7840 Snappy supported
2015/03/26-10:34:35.460682 7f5b322b7840 Zlib supported
2015/03/26-10:34:35.460686 7f5b322b7840 Bzip supported
2015/03/26-10:34:35.460687 7f5b322b7840 LZ4 NOT supported
Reviewers: sdong, yhchiang
Reviewed By: yhchiang
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D35955
Summary:
Improve ThreadStatusSingleCompaction in two ways:
1. Use SYNC_POINT to ensure compaction won't happen
before the test finishes its "Put Phase" instead of
using sleep.
2. In Put Phase, it continues until we have sufficient
number of L0 files. Note that during the put phase,
there won't be any compaction that consumes L0 files
because of item 1.
Test Plan: ./db_test --gtest_filter="*ThreadStatusSingleCompaction*"
Reviewers: sdong, igor, rven
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D35727
Summary:
To understand the bug read t5943287 and check out the new test in column_family_test (ReadDroppedColumnFamily), iter 0.
RocksDB contract allowes you to read a drop column family as long as there is a live reference. However, since our iteration ignores dropped column families, AddLiveFiles() didn't mark files of a dropped column families as live. So we deleted them.
In this patch I no longer ignore dropped column families in the iteration. I think this behavior was confusing and it also led to this bug. Now if an iterator client wants to ignore dropped column families, he needs to do it explicitly.
Test Plan: Added a new unit test that is failing on master. Unit test succeeds now.
Reviewers: sdong, rven, yhchiang
Reviewed By: yhchiang
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D32535
Summary: It is no longer used by the implementation, so we should also remove it from the public API.
Test Plan: make check
Reviewers: sdong
Reviewed By: sdong
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D34971
Summary: Some suggestions for cleanup from Igor.
Test Plan: Regression tests.
Reviewers: igor
Reviewed By: igor
Subscribers: dhruba
Differential Revision: https://reviews.facebook.net/D35169
Summary:
On RocksDB, when there are multiple instances doing
flushes/compactions in the background, the close call takes a long time
because the flushes/compactions need to complete before the database can
shut down. If another instance is using the background threads and the compaction for this instance is in the queue since it has been scheduled, we still cannot shutdown. We now remove the scheduled background tasks which have not yet started running, so that shutdown is speeded up.
Test Plan: DB Test added.
Reviewers: yhchiang, igor, sdong
Reviewed By: sdong
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D33741
Summary:
Here's my proposal for making our LOGs easier to read by machines.
The idea is to dump all events as JSON objects. JSON is easy to read by humans, but more importantly, it's easy to read by machines. That way, we can parse this, load into SQLite/mongo and then query or visualize.
I started with table_create and table_delete events, but if everybody agrees, I'll continue by adding more events (flush/compaction/etc etc)
Test Plan:
Ran db_bench. Observed:
2015/01/15-14:13:25.788019 1105ef000 EVENT_LOG_v1 {"time_micros": 1421360005788015, "event": "table_file_creation", "file_number": 12, "file_size": 1909699}
2015/01/15-14:13:25.956500 110740000 EVENT_LOG_v1 {"time_micros": 1421360005956498, "event": "table_file_deletion", "file_number": 12}
Reviewers: yhchiang, rven, dhruba, MarkCallaghan, lgalanis, sdong
Reviewed By: sdong
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D31647
Summary:
Provide an API which enables users to infor Rocksdb that it is
shutting down.
Test Plan: db_test
Reviewers: sdong, igor
Reviewed By: igor
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D34617
Summary: I want to be able to set this through mongo config.
Test Plan: added unit test
Reviewers: sdong
Reviewed By: sdong
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D34599
Summary:
When having fixed max_bytes_for_level_base, the ratio of size of largest level and the second one can range from 0 to the multiplier. This makes LSM tree frequently irregular and unpredictable. It can also cause poor space amplification in some cases.
In this improvement (proposed by Igor Kabiljo), we introduce a parameter option.level_compaction_use_dynamic_max_bytes. When turning it on, RocksDB is free to pick a level base in the range of (options.max_bytes_for_level_base/options.max_bytes_for_level_multiplier, options.max_bytes_for_level_base] so that real level ratios are close to options.max_bytes_for_level_multiplier.
Test Plan: New unit tests and pass tests suites including valgrind.
Reviewers: MarkCallaghan, rven, yhchiang, igor, ikabiljo
Reviewed By: ikabiljo
Subscribers: yoshinorim, ikabiljo, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D31437
Summary:
When using latest clang (3.6 or 3.7/trunck) rocksdb is failing with many errors. Almost all of them are missing override errors. This diff adds missing override keyword. No manual changes.
Prerequisites: bear and clang 3.5 build with extra tools
```lang=bash
% USE_CLANG=1 bear make all # generate a compilation database http://clang.llvm.org/docs/JSONCompilationDatabase.html
% clang-modernize -p . -include . -add-override
% make format
```
Test Plan:
Make sure all tests are passing.
```lang=bash
% #Use default fb code clang.
% make check
```
Verify less error and no missing override errors.
```lang=bash
% # Have trunk clang present in path.
% ROCKSDB_NO_FBCODE=1 CC=clang CXX=clang++ make
```
Reviewers: igor, kradhakrishnan, rven, meyering, sdong
Reviewed By: sdong
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D34077
Summary:
Prior to this change, "make check" would always waste a lot of
time relinking 60+ binaries. With this change, it does that
only when the generated file, util/build_version.cc, changes,
and that happens only when the date changes or when the
current git SHA changes.
This change makes some other improvements: before, there was no
rule to build a deleted util/build_version.cc. If it was somehow
removed, any attempt to link a program would fail.
There is no longer any need for the separate file,
build_tools/build_detect_version. Its functionality is
now in the Makefile.
* Makefile (DEPFILES): Don't filter-out util/build_version.cc.
No need, and besides, removing that dependency was wrong.
(date, git_sha, gen_build_version): New helper variables.
(util/build_version.cc): New rule, to create this file
and update it only if it would contain new information.
* build_tools/build_detect_platform: Remove file.
* db/db_impl.cc: Now, print only date (not the time).
* util/build_version.h (rocksdb_build_compile_time): Remove
declaration. No longer used.
Test Plan:
- Run "make check" twice, and note that the second time no linking is performed.
- Remove util/build_version.cc and ensure that any "make"
command regenerates it before doing anything else.
- Run this: strings librocksdb.a|grep _build_.
That prints output including the following:
rocksdb_build_git_date:2015-02-19
rocksdb_build_git_sha:2.8.fb-1792-g3cb6cc0
Reviewers: ljin, sdong, igor
Reviewed By: igor
Subscribers: dhruba
Differential Revision: https://reviews.facebook.net/D33591
Summary:
This is a diff for managed iterator. A managed iterator
is a wrapper around an iterator which saves the options for that
iterator as well as the current key/value so that the underlying iterator
and its associated memory can be released when it is aged out
automatically or on the request of the user. Will provide the automatic release as a follow-up diff.
Test Plan: Managed* tests in db_test and XF tests for managed iterator
Reviewers: igor, yhchiang, anthony, sdong
Reviewed By: sdong
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D31401
Summary: Allow GetThreadList to reflect flush activity.
Test Plan:
Developed ThreadStatusFlush test and updated ThreadStatusMultiCompaction test.
./db_test ./thread_list_test
Reviewers: sdong, rven, igor
Reviewed By: igor
Subscribers: dhruba
Differential Revision: https://reviews.facebook.net/D32871
Summary:
It would be good to assing background job their IDs. Two benefits:
1) makes LOGs more readable
2) I might use it in my EventLogger, which will try to make our LOG easier to read/query/visualize
Test Plan: ran rocksdb, read the LOG
Reviewers: sdong, rven, yhchiang
Reviewed By: yhchiang
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D31617
Summary: For description of the bug, see comment in db_test. The fix is pretty straight forward.
Test Plan: added unit test. eventually we need better testing of FOF/POF process.
Reviewers: yhchiang, rven, sdong
Reviewed By: sdong
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D33081
Summary: Having a pointer for DB will be helpful to debug when GDB or working on a dump. If the client process doesn't have any thread actively working on RocksDB, it can be hard to find out.
Test Plan: make all check
Reviewers: rven, yhchiang, igor
Reviewed By: igor
Subscribers: yoshinorim, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D33159
Summary:
This diff basically reverts D30249 and also adds a unit test that was failing before this patch.
I have no idea how I didn't catch this terrible bug when writing a diff, sorry about that :(
I think we should redesign our system of keeping track of and deleting files. This is already a second bug in this critical piece of code. I'll think of few ideas.
BTW this diff is also a regression when running lots of column families. I plan to revisit this separately.
Test Plan: added a unit test
Reviewers: yhchiang, rven, sdong
Reviewed By: sdong
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D33045
Summary:
When DestroyDB() finds a wal file in the DB directory, it assumes it is actually in WAL directory. This can lead to confusion, since it reports IO error when it tries to delete wal file from DB directory. For example: https://ci-builds.fb.com/job/rocksdb_clang_build/296/console
This change will fix our unit tests.
Test Plan: unit tests work
Reviewers: yhchiang, rven, sdong
Reviewed By: sdong
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D32907
Summary:
Add a counter for collecting the wait time on db mutex.
Also add MutexWrapper and CondVarWrapper for measuring wait time.
Test Plan:
./db_test
export ROCKSDB_TESTS=MutexWaitStats
./db_test
verify stats output using db_bench
make clean
make release
./db_bench --statistics=1 --benchmarks=fillseq,readwhilewriting --num=10000 --threads=10
Sample output:
rocksdb.db.mutex.wait.micros COUNT : 7546866
Reviewers: MarkCallaghan, rven, sdong, igor
Reviewed By: igor
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D32787
Summary: This adds a listener for compactions, and gives some useful statistics on each compaction pass.
Test Plan: Unit tests.
Reviewers: sdong, igor, rven, yhchiang
Reviewed By: yhchiang
Subscribers: dhruba
Differential Revision: https://reviews.facebook.net/D31641
Summary: log_dir_unsynced_ is a confusing name. Rename it to log_dir_synced_ and flip the value.
Test Plan: Run ./fault_injection_test
Reviewers: rven, yhchiang, igor
Reviewed By: igor
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D32235
Summary:
1. If WAL directory is different from db directory. Sync the directory after creating a log file under it.
2. After creating an SST file, sync its parent directory instead of DB directory.
3. change the check of kResetDeleteUnsyncedFiles in fault_injection_test. Since we changed the behavior to sync log files' parent directory after first WAL sync, instead of creating, kResetDeleteUnsyncedFiles will not guarantee to show post sync updates.
Test Plan: make all check
Reviewers: yhchiang, rven, igor
Reviewed By: igor
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D32067
Summary: We should not be calling InternalStats methods outside of the mutex.
Test Plan:
COMPILE_WITH_TSAN=1 m db_test && ROCKSDB_TESTS=CompactionTrigger ./db_test
failing before the diff, works now
Reviewers: yhchiang, rven, sdong
Reviewed By: sdong
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D32127
Summary: Now we don't sync manifest file when initializing it, so DB cannot be safely reopened before the first mem table flush. Fix it by syncing it. This fixes fault_injection_test.
Test Plan: make all check
Reviewers: rven, yhchiang, igor
Reviewed By: igor
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D32001
Summary: GetLiveFilesMetaData() already adds a leading "/" in file name. No need to add one extra "/" in DBImpl::CheckConsistency()
Test Plan: make all check
Reviewers: yhchiang, rven, igor
Reviewed By: igor
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D31779
Summary:
This patch remove the unnecessary Compaction::ReleaseInputs().
Compaction::ReleaseInputs() tries to unref its input_version
and column_family. However, such unref is always done in
~Compaction(), and all current ReleaseInputs() calls are
right before the destructor.
Test Plan: ./db_test
Reviewers: igor
Reviewed By: igor
Subscribers: igor, rven, dhruba, sdong
Differential Revision: https://reviews.facebook.net/D31605
Summary:
This patch changes concurrency guarantees around ColumnFamilySet::column_families_ and ColumnFamilySet::column_families_data_.
Before:
* When mutating: lock DB mutex and spin lock
* When reading: lock DB mutex OR spin lock
After:
* When mutating: lock DB mutex and be in write thread
* When reading: lock DB mutex or be in write thread
That way, we eliminate the spin lock that protects these hash maps and simplify concurrency. That means we don't need to lock the spin lock during writing, since writing is mutually exclusive with column family create/drop (the only operations that mutate those hash maps).
With these new restrictions, I also needed to move column family create to the write thread (column family drop was already in the write thread).
Even though we don't need to lock the spin lock during write, impact on performance should be minimal -- the spin lock is almost never busy, so locking it is almost free.
This addresses task t5116919.
Test Plan:
make check
Stress test with lots and lots of column family drop and create:
time ./db_stress --threads=30 --ops_per_thread=5000000 --max_key=5000 --column_families=200 --clear_column_family_one_in=100000 --verify_before_write=0 --reopen=15 --max_background_compactions=10 --max_background_flushes=10 --db=/fast-rocksdb-tmp/db_stress/
Reviewers: yhchiang, rven, sdong
Reviewed By: sdong
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D30651
