Summary:
The way DBImpl::TEST_CompactRange() throttles down the number of bg compactions
can cause it to deadlock when CompactRange() is called concurrently from
multiple threads. Imagine a following scenario with only two threads
(max_background_compactions is 10 and bg_compaction_scheduled_ is initially 0):
1. Thread #1 increments bg_compaction_scheduled_ (to LargeNumber), sets
bg_compaction_scheduled_ to 9 (newvalue), schedules the compaction
(bg_compaction_scheduled_ is now 10) and waits for it to complete.
2. Thread #2 calls TEST_CompactRange(), increments bg_compaction_scheduled_
(now LargeNumber + 10) and waits on a cv for bg_compaction_scheduled_ to
drop to LargeNumber.
3. BG thread completes the first manual compaction, decrements
bg_compaction_scheduled_ and wakes up all threads waiting on bg_cv_.
Thread #1 runs, increments bg_compaction_scheduled_ by LargeNumber again
(now 2*LargeNumber + 9). Since that's more than LargeNumber + newvalue,
thread #2 also goes to sleep (waiting on bg_cv_), without resetting
bg_compaction_scheduled_.
This diff attempts to address the problem by introducing a new counter
bg_manual_only_ (when positive, MaybeScheduleFlushOrCompaction() will only
schedule manual compactions).
Test Plan:
I could pretty much consistently reproduce the deadlock with a program that
calls CompactRange(nullptr, nullptr) immediately after Write() from multiple
threads. This no longer happens with this patch.
Tests (make check) pass.
Reviewers: dhruba, igor, sdong, haobo
Reviewed By: igor
CC: leveldb
Differential Revision: https://reviews.facebook.net/D14799
Summary:
The rule file is forked from that in Facebook's repo.
I'll add format file for now and team members can tune the rules later.
In this patch, I made only two changes in order to be consistent with existing coding style
`SpacesBeforeTrailingComments: 2`
`ColumnLimit: 80`
Test Plan: N/A
Reviewers: igor
CC: leveldb
Differential Revision: https://reviews.facebook.net/D15015
Summary: Added a script that prepares the repo for facebook's new rocksdb release, which will automatically do some necessary work to make sure this repo is ready for 3rdparty release.
Test Plan:
Run this script and observed:
* new version was created (both in local and remote repo) as a git tag.
* build_version.cc was updated
* build_detect_platform was changed so that it won't create any new change.
Reviewers: haobo, dhruba, sdong, igor
CC: leveldb
Differential Revision: https://reviews.facebook.net/D15003
Summary:
We don't want two threads to clash if they concurrently call DisableFileDeletions() and EnableFileDeletions(). I'm adding a counter that will enable file deletions only after all DisableFileDeletions() calls have been negated with EnableFileDeletions().
However, we also don't want to break the old behavior, so I added a parameter force to EnableFileDeletions(). If force is true, we will still enable file deletions after every call to EnableFileDeletions(), which is what is happening now.
Test Plan: make check
Reviewers: dhruba, haobo, sanketh
Reviewed By: dhruba
CC: leveldb
Differential Revision: https://reviews.facebook.net/D14781
Summary: I'm not sure what's the purpose of encoding file number to a new buffer for looking up the table cache. It seems to be unnecessary to me. With this patch, we point the lookup key to the address of the int64 of the file number.
Test Plan: make all check
Reviewers: dhruba, haobo, igor, kailiu
Reviewed By: dhruba
CC: leveldb
Differential Revision: https://reviews.facebook.net/D14811
Summary:
In some places we have NotFound status created with empty message, but it doesn't avoid a malloc. With this patch, the malloc is avoided for that case.
The motivation of it is that I found in db_bench readrandom test when all keys are not existing, about 4% of the total running time is spent on malloc of Status, plus a similar amount of CPU spent on free of them, which is not necessary.
Test Plan: make all check
Reviewers: dhruba, haobo, igor
Reviewed By: haobo
CC: leveldb
Differential Revision: https://reviews.facebook.net/D14691
Summary:
A vector that leverages pre-allocated stack-based array to achieve better
performance for array with small amount of items.
Test Plan:
Added tests for both correctness and performance
Here is the performance benchmark between vector and autovector
Please note that in the test "Creation and Insertion Test", the test case were designed with the motivation described below:
* no element inserted: internal array of std::vector may not really get
initialize.
* one element inserted: internal array of std::vector must have
initialized.
* kSize elements inserted. This shows the most time we'll spend if we
keep everything in stack.
* 2 * kSize elements inserted. The internal vector of
autovector must have been initialized.
Note: kSize is the capacity of autovector
=====================================================
Creation and Insertion Test
=====================================================
created 100000 vectors:
each was inserted with 0 elements
total time elapsed: 128000 (ns)
created 100000 autovectors:
each was inserted with 0 elements
total time elapsed: 3641000 (ns)
created 100000 VectorWithReserveSizes:
each was inserted with 0 elements
total time elapsed: 9896000 (ns)
-----------------------------------
created 100000 vectors:
each was inserted with 1 elements
total time elapsed: 11089000 (ns)
created 100000 autovectors:
each was inserted with 1 elements
total time elapsed: 5008000 (ns)
created 100000 VectorWithReserveSizes:
each was inserted with 1 elements
total time elapsed: 24271000 (ns)
-----------------------------------
created 100000 vectors:
each was inserted with 4 elements
total time elapsed: 39369000 (ns)
created 100000 autovectors:
each was inserted with 4 elements
total time elapsed: 10121000 (ns)
created 100000 VectorWithReserveSizes:
each was inserted with 4 elements
total time elapsed: 28473000 (ns)
-----------------------------------
created 100000 vectors:
each was inserted with 8 elements
total time elapsed: 75013000 (ns)
created 100000 autovectors:
each was inserted with 8 elements
total time elapsed: 18237000 (ns)
created 100000 VectorWithReserveSizes:
each was inserted with 8 elements
total time elapsed: 42464000 (ns)
-----------------------------------
created 100000 vectors:
each was inserted with 16 elements
total time elapsed: 102319000 (ns)
created 100000 autovectors:
each was inserted with 16 elements
total time elapsed: 76724000 (ns)
created 100000 VectorWithReserveSizes:
each was inserted with 16 elements
total time elapsed: 68285000 (ns)
-----------------------------------
=====================================================
Sequence Access Test
=====================================================
performed 100000 sequence access against vector
size: 4
total time elapsed: 198000 (ns)
performed 100000 sequence access against autovector
size: 4
total time elapsed: 306000 (ns)
-----------------------------------
performed 100000 sequence access against vector
size: 8
total time elapsed: 565000 (ns)
performed 100000 sequence access against autovector
size: 8
total time elapsed: 512000 (ns)
-----------------------------------
performed 100000 sequence access against vector
size: 16
total time elapsed: 1076000 (ns)
performed 100000 sequence access against autovector
size: 16
total time elapsed: 1070000 (ns)
-----------------------------------
Reviewers: dhruba, haobo, sdong, chip
Reviewed By: dhruba
CC: leveldb
Differential Revision: https://reviews.facebook.net/D14655
Summary:
Instead of locking and saving a DB state, we can cache a DB state and update it only when it changes. This change reduces lock contention and speeds up read operations on the DB.
Performance improvements are substantial, although there is some cost in no-read workloads. I ran the regression tests on my devserver and here are the numbers:
overwrite 56345 -> 63001
fillseq 193730 -> 185296
readrandom 771301 -> 1219803 (58% improvement!)
readrandom_smallblockcache 677609 -> 862850
readrandom_memtable_sst 710440 -> 1109223
readrandom_fillunique_random 221589 -> 247869
memtablefillrandom 105286 -> 92643
memtablereadrandom 763033 -> 1288862
Test Plan:
make asan_check
I am also running db_stress
Reviewers: dhruba, haobo, sdong, kailiu
Reviewed By: haobo
CC: leveldb
Differential Revision: https://reviews.facebook.net/D14679
Summary:
For some tests I want to cache the database prior to running other tests on the same invocation
of db_bench. The readtocache test ignores --threads and --reads so those can be used by other tests
and it will still do a full read of --num rows with one thread. It might be invoked like:
db_bench --benchmarks=readtocache,readrandom --reads 100 --num 10000 --threads 8
Task ID: #
Blame Rev:
Test Plan:
run db_bench
Revert Plan:
Database Impact:
Memcache Impact:
Other Notes:
EImportant:
- begin *PUBLIC* platform impact section -
Bugzilla: #
- end platform impact -
Reviewers: dhruba
CC: leveldb
Differential Revision: https://reviews.facebook.net/D14739
Summary:
db_test should be the first to execute because it finds the most bugs.
Also, when third parties report issues, we don't want ldb error message, we prefer to have db_test error message. For example, see thread: https://github.com/facebook/rocksdb/issues/25
Test Plan: make check
Reviewers: dhruba, haobo, kailiu
Reviewed By: dhruba
CC: leveldb
Differential Revision: https://reviews.facebook.net/D14715
Summary: I realized that "D14409 Avoid sorting in Version::Get() by presorting them in VersionSet::Builder::SaveTo()" is not done in an optimized place. SaveTo() is usually inside mutex. Move it to Finalize(), which is called out of mutex.
Test Plan: make all check
Reviewers: dhruba, haobo, kailiu
Reviewed By: dhruba
CC: igor, leveldb
Differential Revision: https://reviews.facebook.net/D14607
Summary: It seems to be a decision tradeoff in current codes: we make a malloc for every Get() to reduce one malloc for a flush inside mutex. It takes about 5% of CPU time in readrandom tests. We might consider the tradeoff to be the other way around.
Test Plan: make all check
Reviewers: dhruba, haobo, igor
Reviewed By: dhruba
CC: leveldb
Differential Revision: https://reviews.facebook.net/D14697
Summary:
I realized that manifest will get deleted by PurgeObsoleteFiles in DBImpl, but it is sill cleaner to delete
files before we restore the backup
Test Plan: backupable_db_test
Reviewers: dhruba
Reviewed By: dhruba
CC: leveldb
Differential Revision: https://reviews.facebook.net/D14619
Summary: @MarkCallaghan's tests indicate that performance with 8k rows in memtable is much worse than empty memtable. I wanted to add a regression tests that measures this effect, so we could optimize it. However, current config shows 634461 QPS on my devbox. Mark, any idea why this is so much faster than your measurements?
Test Plan: Ran the regression test.
Reviewers: MarkCallaghan, dhruba, haobo
Reviewed By: MarkCallaghan
CC: leveldb, MarkCallaghan
Differential Revision: https://reviews.facebook.net/D14511
Summary: In get operations, merge_operands is only used in few cases. Lazily initialize it can reduce average latency in some cases
Test Plan: make all check
Reviewers: haobo, kailiu, dhruba
Reviewed By: haobo
CC: igor, nkg-, leveldb
Differential Revision: https://reviews.facebook.net/D14415
Conflicts:
db/db_impl.cc
db/memtable.cc
Summary: Pre-sort files in VersionSet::Builder::SaveTo() so that when getting the value, no need to sort them. It can avoid the costs of vector operations and sorting in Version::Get().
Test Plan: make all check
Reviewers: haobo, kailiu, dhruba
Reviewed By: dhruba
CC: nkg-, igor, leveldb
Differential Revision: https://reviews.facebook.net/D14409
Summary:
creating new iterators of mem tables can be expensive. Move them out of mutex.
DBImpl::WriteLevel0Table()'s mems seems to be a local vector and is only used by flushing. memtables to flush are also immutable, so it should be safe to do so.
Test Plan: make all check
Reviewers: haobo, dhruba, kailiu
Reviewed By: dhruba
CC: igor, leveldb
Differential Revision: https://reviews.facebook.net/D14577
Conflicts:
db/db_impl.cc
Summary:
I have ran a get benchmark where all the data is in the cache and observed that most of the time is spent on waiting for lock in LRUCache.
This is an effort to optimize LRUCache.
Test Plan:
The data was loaded with fillseq. Then, I ran a benchmark:
/db_bench --db=/tmp/rocksdb_stat_bench --num=1000000 --benchmarks=readrandom --statistics=1 --use_existing_db=1 --threads=16 --disable_seek_compaction=1 --cache_size=20000000000 --cache_numshardbits=8 --table_cache_numshardbits=8
I ran the benchmark three times. Here are the results:
AFTER THE PATCH: 798072, 803998, 811807
BEFORE THE PATCH: 782008, 815593, 763017
Reviewers: dhruba, haobo, kailiu
Reviewed By: haobo
CC: leveldb
Differential Revision: https://reviews.facebook.net/D14571
Summary: We now delete backups with newer sequence number, so the clients don't have to handle confusing situations when they restore from backup.
Test Plan: added a unit test
Reviewers: dhruba
Reviewed By: dhruba
CC: leveldb
Differential Revision: https://reviews.facebook.net/D14547
Tomislav Novak