Summary:
This diff allows a Writer to join the next write batch group
without acquiring any locks. Waiting is performed via a per-Writer mutex,
so all of the non-leader writers never need to acquire the db mutex.
It is now possible to join a write batch group after the leader has been
chosen but before the batch has been constructed. This diff doesn't
increase parallelism, but reduces synchronization overheads.
For some CPU-bound workloads (no WAL, RAM-sized working set) this can
substantially reduce contention on the db mutex in a multi-threaded
environment. With T=8 N=500000 in a CPU-bound scenario (see the test
plan) this is good for a 33% perf win. Not all scenarios see such a
win, but none show a loss. This code is slightly faster even for the
single-threaded case (about 2% for the CPU-bound scenario below).
Test Plan:
1. unit tests
2. COMPILE_WITH_TSAN=1 make check
3. stress high-contention scenarios with db_bench -benchmarks=fillrandom -threads=$T -batch_size=1 -memtablerep=skip_list -value_size=0 --num=$N -level0_slowdown_writes_trigger=9999 -level0_stop_writes_trigger=9999 -disable_auto_compactions --max_write_buffer_number=8 -max_background_flushes=8 --disable_wal --write_buffer_size=160000000
Reviewers: sdong, igor, rven, ljin, yhchiang
Subscribers: dhruba
Differential Revision: https://reviews.facebook.net/D43887
Summary:
Add options.compaction_measure_io_stats to print out / pass to listener accumulated time spent on write calls. Example outputs in info logs:
2015/08/12-16:27:59.463944 7fd428bff700 (Original Log Time 2015/08/12-16:27:59.463922) EVENT_LOG_v1 {"time_micros": 1439422079463897, "job": 6, "event": "compaction_finished", "output_level": 1, "num_output_files": 4, "total_output_size": 6900525, "num_input_records": 111483, "num_output_records": 106877, "file_write_nanos": 15663206, "file_range_sync_nanos": 649588, "file_fsync_nanos": 349614797, "file_prepare_write_nanos": 1505812, "lsm_state": [2, 4, 0, 0, 0, 0, 0]}
Add two more counters in iostats_context.
Also add a parameter of db_bench.
Test Plan: Add a unit test. Also manually verify LOG outputs in db_bench
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D44115
Summary: Iterator has a bug: if a child iterator reaches its end, and user issues a Prev(), and just before SeekToLast() of the child iterator is called, some extra rows is added in the end, the position of iterator can be misplaced.
Test Plan: Run the tests with or without valgrind
Reviewers: rven, yhchiang, IslamAbdelRahman, anthony
Reviewed By: anthony
Subscribers: tnovak, leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D43671
Summary:
Initial implementation of Pessimistic Transactions. This diff contains the api changes discussed in D38913. This diff is pretty large, so let me know if people would prefer to meet up to discuss it.
MyRocks folks: please take a look at the API in include/rocksdb/utilities/transaction[_db].h and let me know if you have any issues.
Also, you'll notice a couple of TODOs in the implementation of RollbackToSavePoint(). After chatting with Siying, I'm going to send out a separate diff for an alternate implementation of this feature that implements the rollback inside of WriteBatch/WriteBatchWithIndex. We can then decide which route is preferable.
Next, I'm planning on doing some perf testing and then integrating this diff into MongoRocks for further testing.
Test Plan: Unit tests, db_bench parallel testing.
Reviewers: igor, rven, sdong, yhchiang, yoshinorim
Reviewed By: sdong
Subscribers: hermanlee4, maykov, spetrunia, leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D40869
Summary: Add a new option that all LoadTableHandlers to use multiple threads to load files on DB Open and Recover
Test Plan:
make check -j64
COMPILE_WITH_TSAN=1 make check -j64
DISABLE_JEMALLOC=1 make all valgrind_check -j64 (still running)
Reviewers: yhchiang, anthony, rven, kradhakrishnan, igor, sdong
Reviewed By: sdong
Subscribers: dhruba
Differential Revision: https://reviews.facebook.net/D43755
Summary:
While working on single delete support for db_bench, I realized that
db_bench/db_stress contain a bunch of duplicate code related to
copmression and found some typos. This patch removes duplicate code,
typos and a redundant #ifndef in internal_stats.cc.
Test Plan: make db_stress && make db_bench && ./db_bench --benchmarks=compress,uncompress
Reviewers: yhchiang, sdong, rven, anthony, igor
Reviewed By: igor
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D43965
Summary:
Key comparison is the single largest CPU user for CPU-bound
workloads. This diff reduces the number of comparisons in two ways.
The first is that it moves predecessor array gathering from
FindGreaterOrEqual to FindLessThan, so that FindGreaterOrEqual can
return immediately if compare_ returns 0. As part of this change I
moved the sequential insertion optimization into Insert, to remove the
undocumented (and smelly) requirement that prev must be equal to prev_
if it is non-null.
The second optimization is that all of the search functions skip calling
compare_ when moving to a lower level that has the same Next pointer.
With a branching factor of 4 we would expect this to happen 1/4 of
the time.
On a single-threaded CPU-bound workload (-benchmarks=fillrandom -threads=1
-batch_size=1 -memtablerep=skip_list -value_size=0 --num=1600000
-level0_slowdown_writes_trigger=9999 -level0_stop_writes_trigger=9999
-disable_auto_compactions --max_write_buffer_number=8
-max_background_flushes=8 --disable_wal --write_buffer_size=160000000)
on my dev server this is good for a 7% perf win.
Test Plan: unit tests
Reviewers: rven, ljin, yhchiang, sdong, igor
Reviewed By: igor
Subscribers: dhruba
Differential Revision: https://reviews.facebook.net/D43233
Summary:
Changed compaction_job_test to support better/more thorough
tests and added two tests. Also changed MockFileContents
to order using InternalKeyComparator.
Test Plan: make compaction_job_test && ./compaction_job_test; make all && make check
Reviewers: sdong, rven, igor, yhchiang, anthony
Reviewed By: anthony
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D42837
Summary: Implemented this simple wrapper for something else I was working on. Seemed like it makes sense to expose it instead of burying it in some random code.
Test Plan: added test
Reviewers: rven, kradhakrishnan, sdong, yhchiang
Reviewed By: sdong
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D43293
Summary:
Visual Studio complains about deque<LogWriterNumber> because LogWriterNumber is non-copyable for its unique_ptr member writer. Move away from it, and do explit free.
It is less safe but I can't think of a better way to unblock it.
Test Plan: valgrind check test
Reviewers: anthony, IslamAbdelRahman, kolmike, rven, yhchiang
Reviewed By: yhchiang
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D43647
Summary:
When seeking to the last occurrence of a key with sequence number 0, db_iter
ends up in an endless loop because it seeks to type kValueTypeForSeek
which is larger than kTypeDeletion/kTypeValue. Added test case that triggers
the behavior.
Test Plan: make clean all check
Reviewers: igor, rven, anthony, yhchiang, sdong
Reviewed By: sdong
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D43653
Summary: Update DeleteScheduler tests so that they verify the used penalties for waiting instead of measuring the time spent which is not reliable
Test Plan:
make -j64 delete_scheduler_test && ./delete_scheduler_test
COMPILE_WITH_TSAN=1 make -j64 delete_scheduler_test && ./delete_scheduler_test
COMPILE_WITH_ASAN=1 make -j64 delete_scheduler_test && ./delete_scheduler_test
make -j64 db_test && ./db_test --gtest_filter="DBTest.RateLimitedDelete:DBTest.DeleteSchedulerMultipleDBPaths"
COMPILE_WITH_TSAN=1 make -j64 db_test && ./db_test --gtest_filter="DBTest.RateLimitedDelete:DBTest.DeleteSchedulerMultipleDBPaths"
COMPILE_WITH_ASAN=1 make -j64 db_test && ./db_test --gtest_filter="DBTest.RateLimitedDelete:DBTest.DeleteSchedulerMultipleDBPaths"
Reviewers: yhchiang, sdong
Reviewed By: sdong
Subscribers: dhruba
Differential Revision: https://reviews.facebook.net/D43635
Summary: The list of info log files of a db can be obtained using the new function.
Test Plan: New test in db_test.cc passed.
Reviewers: yhchiang, IslamAbdelRahman, sdong
Reviewed By: sdong
Subscribers: IslamAbdelRahman, leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D41715
Summary:
Add two unit tests for SyncWAL(). One makes sure SyncWAL() doesn't block writes in the other thread. Another one makes sure SyncWAL() doesn't wait ongoing writes to finish before being executed.
Create a new test file db_wal_test and move two WAL related tests from db_test to here.
Test Plan: Run the new tests
Reviewers: IslamAbdelRahman, rven, kradhakrishnan, kolmike, tnovak, yhchiang
Reviewed By: yhchiang
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D43605
Summary: Measure read latency histogram and put in statistics. Compaction inputs are excluded from it when possible (unfortunately usually no possible as we usually take table reader from table cache.
Test Plan:
Run db_bench and it shows the stats, like:
rocksdb.sst.read.micros statistics Percentiles :=> 50 : 1.238522 95 : 2.529740 99 : 3.912180
Reviewers: kradhakrishnan, rven, anthony, IslamAbdelRahman, MarkCallaghan, yhchiang
Reviewed By: yhchiang
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D43275
Summary: This patch will fix the false positive of DBTest.FlushSchedule under TSAN, we dont need to disable this test
Test Plan: COMPILE_WITH_TSAN=1 make -j64 db_test && ./db_test --gtest_filter="DBTest.FlushSchedule"
Reviewers: yhchiang, sdong
Reviewed By: sdong
Subscribers: dhruba
Differential Revision: https://reviews.facebook.net/D43599
Summary:
While doing forward iterating, if current key is merge, internal iterator position is placed to the next key. If Prev() is called now, needs to do extra Prev() to recover the location.
This is second attempt of fixing after reverting ec70fea4c4. This time shrink the fix to only merge key is the current key and avoid the reseeking logic for max_iterating skipping
Test Plan: enable the two disabled tests and make sure they pass
Reviewers: rven, IslamAbdelRahman, kradhakrishnan, tnovak, yhchiang
Reviewed By: yhchiang
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D43557
Summary:
While working on https://reviews.facebook.net/D43179 , I found
duplicate code in the tests. This patch removes it.
Test Plan: make clean all check
Reviewers: igor, sdong, rven, anthony, yhchiang
Reviewed By: yhchiang
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D43263
Summary:
Subj. We really need this feature.
Previous diff D40899 has most of the changes to make this possible, this diff just adds the method.
Test Plan: `make check`, the new test fails without this diff; ran with ASAN, TSAN and valgrind.
Reviewers: igor, rven, IslamAbdelRahman, anthony, kradhakrishnan, tnovak, yhchiang, sdong
Reviewed By: sdong
Subscribers: MarkCallaghan, maykov, hermanlee4, yoshinorim, tnovak, dhruba
Differential Revision: https://reviews.facebook.net/D40905
Summary:
Updated DBTest DBCompactionTest and CompactionJobStatsTest
to run compaction-related tests once with subcompactions enabled and
once disabled using the TEST_P test type in the Google Test suite.
Test Plan: ./db_test ./db_compaction-test ./compaction_job_stats_test
Reviewers: sdong, igor, anthony, yhchiang
Reviewed By: yhchiang
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D43443
Summary:
Introduce DeleteScheduler that allow enforcing a rate limit on file deletion
Instead of deleting files immediately, files are moved to trash directory and deleted in a background thread that apply sleep penalty between deletes if needed.
I have updated PurgeObsoleteFiles and PurgeObsoleteWALFiles to use the delete_scheduler instead of env_->DeleteFile
Test Plan:
added delete_scheduler_test
existing unit tests
Reviewers: kradhakrishnan, anthony, rven, yhchiang, sdong
Reviewed By: sdong
Subscribers: dhruba
Differential Revision: https://reviews.facebook.net/D43221
Summary:
Make DBCompactionTest.SkipStatsUpdateTest more stable by
removing flaky but unnecessary assertion on the size of db
as simply checking the random file open count is suffice.
Test Plan: db_compaction_test
Reviewers: igor, anthony, IslamAbdelRahman, sdong
Reviewed By: sdong
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D43533
Summary:
UpdateAccumulatedStats() is used to optimize compaction decision
esp. when the number of deletion entries are high, but this function
can slowdown DBOpen esp. in disk environment.
This patch adds DBOptions::skip_sats_update_on_db_open, which skips
UpdateAccumulatedStats() in DB::Open() time when it's set to true.
Test Plan: Add DBCompactionTest.SkipStatsUpdateTest
Reviewers: igor, anthony, IslamAbdelRahman, sdong
Reviewed By: sdong
Subscribers: tnovak, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D42843
Summary:
The compact files API had a bug where some overlapping files
are not added. These are files which overlap with files which were
added to the compaction input files, but not to the original set of
input files. This happens only when there are more than two levels
involved in the compaction. An example will illustrate this better.
Level 2 has 1 input file 1.sst which spans [20,30].
Level 3 has added file 2.sst which spans [10,25]
Level 4 has file 3.sst which spans [35,40] and
input file 4.sst which spans [46,50].
The existing code would not add 3.sst to the set of input_files because
it only becomes an overlapping file in level 4 and it wasn't one in
level 3.
When installing the results of the compaction, 3.sst would overlap with
output file from the compact files and result in the assertion in
version_set.cc:1130
// Must not overlap
assert(level <= 0 || level_files->empty() ||
internal_comparator_->Compare(
(*level_files)[level_files->size() - 1]->largest, f->smallest) <
0);
This change now adds overlapping files from the current level to the set
of input files also so that we don't hit the assertion above.
Test Plan:
d=/tmp/j; rm -rf $d; seq 1000 | parallel --gnu --eta
'd=/tmp/j/d-{}; mkdir -p $d; TEST_TMPDIR=$d ./db_compaction_test
--gtest_filter=*CompactilesOnLevel* --gtest_also_run_disabled_tests >&
'$d'/log-{}'
Reviewers: igor, yhchiang, sdong
Reviewed By: yhchiang
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D43437
Summary:
Made SuggestCompactRangeNoTwoLevel0Compactions by forcing
a flush after generating a file and waiting for compaction at the end.
Test Plan: Run SuggestCompactRangeNoTwoLevel0Compactions
Reviewers: yhchiang, igor, sdong
Reviewed By: sdong
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D43449
Summary:
As of now compactions involving files from Level 0 and Level 1 are single
threaded because the files in L0, although sorted, are not range partitioned like
the other levels. This means that during L0-L1 compaction each file from L1
needs to be merged with potentially all the files from L0.
This attempt to parallelize the L0-L1 compaction assigns a thread and a
corresponding iterator to each L1 file that then considers only the key range
found in that L1 file and only the L0 files that have those keys (and only the
specific portion of those L0 files in which those keys are found). In this way
the overlap is minimized and potentially eliminated between different iterators
focusing on the same files.
The first step is to restructure the compaction logic to break L0-L1 compactions
into multiple, smaller, sequential compactions. Eventually each of these smaller
jobs will be run simultaneously. Areas to pay extra attention to are
# Correct aggregation of compaction job statistics across multiple threads
# Proper opening/closing of output files (make sure each thread's is unique)
# Keys that span multiple L1 files
# Skewed distributions of keys within L0 files
Test Plan: Make and run db_test (newer version has separate compaction tests) and compaction_job_stats_test
Reviewers: igor, noetzli, anthony, sdong, yhchiang
Reviewed By: yhchiang
Subscribers: MarkCallaghan, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D42699
Summary:
There was a bug in table_properties_collector_test that this patch
is fixing: `!backward_mode && !test_int_tbl_prop_collector` in
TestCustomizedTablePropertiesCollector was never true, so the code
in the if-block never got executed. The reason is that the
CustomizedTablePropertiesCollector test was skipping tests with
`!backward_mode_ && !encode_as_internal`. The reason for skipping
the tests is unknown.
Test Plan: make table_properties_collector_test && ./table_properties_collector_test
Reviewers: rven, igor, yhchiang, anthony, sdong
Reviewed By: sdong
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D43281
Summary:
Support RollbackToSavePoint() in WriteBatch and WriteBatchWithIndex. Support for partial transaction rollback is needed for MyRocks.
An alternate implementation of Transaction::RollbackToSavePoint() exists in D40869. However, the other implementation is messier because it is implemented outside of WriteBatch. This implementation is much cleaner and also exposes a potentially useful feature to WriteBatch.
Test Plan: Added unit tests
Reviewers: IslamAbdelRahman, kradhakrishnan, maykov, yoshinorim, hermanlee4, spetrunia, sdong, yhchiang
Reviewed By: yhchiang
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D42723
Summary:
For task #7771355, we would like to log the number of corrupt keys
during a compaction. This patch implements and tests the count
as part of CompactionJobStats.
Test Plan: make && make check
Reviewers: rven, igor, yhchiang, sdong
Reviewed By: sdong
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D42921
Summary: Fix for universal compaction with trivial move, when the ouput level is 0. The tests where failing. Fixed by allowing normal compaction when output level is 0.
Test Plan: modified test cases run successfully.
Reviewers: sdong, yhchiang, IslamAbdelRahman
Reviewed By: IslamAbdelRahman
Subscribers: anthony, kradhakrishnan, leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D42933
Summary: DBCompactionTest.PartialCompactionFailure has a risk that one flush job writes out two mem tables into one file, so that the total files flushed are less than expected. Fix it by writing for flush to finish after every write.
Test Plan: Run the test
Reviewers: IslamAbdelRahman, kradhakrishnan, yhchiang, anthony
Reviewed By: anthony
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D42831
Summary: These tests used to fail if a compaction happened between flushing tables and enumerating them to get properties.
Test Plan: this reports occasional failures without this diff and no failures with it: `for i in {1..10000}; do echo $i; done | parallel --gnu -j100 'TEST_TMPDIR=`TMPDIR=/dev/shm/rockstemp mktemp -d -t` ./db_test --gtest_filter=DBTest.GetUserDefinedTablaProperties >&/dev/null || echo {} failed'`
Reviewers: sdong
Reviewed By: sdong
Subscribers: dhruba
Differential Revision: https://reviews.facebook.net/D42861
Summary:
I'll just copy internal task summary here:
"
This sequence will cause data loss in the middle after an sync write:
non-sync write key 1
flush triggered, not yet scheduled
sync write key 2
system crash
After rebooting, users might see key 2 but not key 1, which violates the API of sync write.
This can be reproduced using unit test FaultInjectionTest::DISABLED_WriteOptionSyncTest.
One way to fix it is for a sync write, if there is outstanding unsynced log files, we need to syc them too.
"
This diff should be considered together with the next diff D40905; in isolation this fix probably could be a little simpler.
Test Plan: `make check`; added a test for that (DBTest.SyncingPreviousLogs) before noticing FaultInjectionTest.WriteOptionSyncTest (keeping both since mine asserts a bit more); both tests fail without this diff; for D40905 stacked on top of this diff, ran tests with ASAN, TSAN and valgrind
Reviewers: rven, yhchiang, IslamAbdelRahman, anthony, kradhakrishnan, igor, sdong
Reviewed By: sdong
Subscribers: dhruba
Differential Revision: https://reviews.facebook.net/D40899
Summary:
Fixes T6548822. Added a new function for estimating the size of the live data
as proposed in the task. The value can be accessed through the property
rocksdb.estimate-live-data-size.
Test Plan:
There are two unit tests in version_set_test and a simple test in db_test.
make version_set_test && ./version_set_test;
make db_test && ./db_test gtest_filter=GetProperty
Reviewers: rven, igor, yhchiang, sdong
Reviewed By: sdong
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D41493
Summary: DBTest.GetPropertiesOfAllTablesTest generates four files and expects four files there, but a L0->L1 comapction can trigger to compact to one single file. Fix it by raising level 0 number of file compaction trigger
Test Plan: Run it many times and see it never fails.
Reviewers: kradhakrishnan, IslamAbdelRahman, yhchiang, anthony
Reviewed By: anthony
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D42789
Summary: Move general compaction tests from db_test.cc to db_compaction_test.cc
Test Plan:
db_test
db_compaction_test
Reviewers: igor, sdong, IslamAbdelRahman, anthony
Reviewed By: anthony
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D42651
Summary: Add new CheckFileExists method. Considered changing the FileExists api but didn't want to break anyone's builds.
Test Plan: unit tests
Reviewers: yhchiang, igor, sdong
Reviewed By: sdong
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D42003
Summary:
Fixed DBTestUniversalManualCompactionOutputPathId test
by changing the expected number of files when setting up
the test as flushes no-longer preempt compactions
in patch https://reviews.facebook.net/D41931.
Also, include db_universal_copaction_test in make all check.
Test Plan: db_universal_copaction_test
Reviewers: igor, sdong, IslamAbdelRahman, anthony
Reviewed By: anthony
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D42639
Summary: This unit test is blocking our release since it fails under certain
compiler versions. The failure is due to a race in the unit test and not the
core functionality.
Test Plan: Run locally
Reviewers: sdong
CC: leveldb
Task ID: #7760955
Blame Rev:
Summary: Now we allow trivial move in universal compaction. Add a parameter in db_bench
Test Plan: Run db_bench with this option on and off and make sure the option is switched correctly.
Reviewers: yhchiang, igor, kradhakrishnan, anthony
Reviewed By: anthony
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D41427
Summary: Now the major test cases of fault_injection_test only insert keys in sorted order so compactions will be trivial move. Add a new mode to insert in non-sequential order to trigger non-trivial compactions.
Test Plan: Run the test
Reviewers: kradhakrishnan, anthony, IslamAbdelRahman, yhchiang, igor
Reviewed By: igor
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D42435
Summary:
Skipping these tests in ROCKSDB_LITE since they are not supported
json_document_test
wal_manager_test
ttl_test
sst_dump_test
deletefile_test
compact_files_test
prefix_test
checkpoint_test
Test Plan:
json_document_test
wal_manager_test
ttl_test
sst_dump_test
deletefile_test
compact_files_test
prefix_test
checkpoint_test
Reviewers: igor, sdong, yhchiang, kradhakrishnan, anthony
Reviewed By: anthony
Subscribers: dhruba
Differential Revision: https://reviews.facebook.net/D42573
Summary: Make merge_test runnable in ROCKSDB_LITE
Test Plan: merge_test
Reviewers: igor, sdong, yhchiang, kradhakrishnan, anthony
Reviewed By: anthony
Subscribers: dhruba
Differential Revision: https://reviews.facebook.net/D42579
