Summary:
This diff adds support for concurrent adds to the skiplist memtable
implementations. Memory allocation is made thread-safe by the addition of
a spinlock, with small per-core buffers to avoid contention. Concurrent
memtable writes are made via an additional method and don't impose a
performance overhead on the non-concurrent case, so parallelism can be
selected on a per-batch basis.
Write thread synchronization is an increasing bottleneck for higher levels
of concurrency, so this diff adds --enable_write_thread_adaptive_yield
(default off). This feature causes threads joining a write batch
group to spin for a short time (default 100 usec) using sched_yield,
rather than going to sleep on a mutex. If the timing of the yield calls
indicates that another thread has actually run during the yield then
spinning is avoided. This option improves performance for concurrent
situations even without parallel adds, although it has the potential to
increase CPU usage (and the heuristic adaptation is not yet mature).
Parallel writes are not currently compatible with
inplace updates, update callbacks, or delete filtering.
Enable it with --allow_concurrent_memtable_write (and
--enable_write_thread_adaptive_yield). Parallel memtable writes
are performance neutral when there is no actual parallelism, and in
my experiments (SSD server-class Linux and varying contention and key
sizes for fillrandom) they are always a performance win when there is
more than one thread.
Statistics are updated earlier in the write path, dropping the number
of DB mutex acquisitions from 2 to 1 for almost all cases.
This diff was motivated and inspired by Yahoo's cLSM work. It is more
conservative than cLSM: RocksDB's write batch group leader role is
preserved (along with all of the existing flush and write throttling
logic) and concurrent writers are blocked until all memtable insertions
have completed and the sequence number has been advanced, to preserve
linearizability.
My test config is "db_bench -benchmarks=fillrandom -threads=$T
-batch_size=1 -memtablerep=skip_list -value_size=100 --num=1000000/$T
-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
--block_size=16384 --allow_concurrent_memtable_write" on a two-socket
Xeon E5-2660 @ 2.2Ghz with lots of memory and an SSD hard drive. With 1
thread I get ~440Kops/sec. Peak performance for 1 socket (numactl
-N1) is slightly more than 1Mops/sec, at 16 threads. Peak performance
across both sockets happens at 30 threads, and is ~900Kops/sec, although
with fewer threads there is less performance loss when the system has
background work.
Test Plan:
1. concurrent stress tests for InlineSkipList and DynamicBloom
2. make clean; make check
3. make clean; DISABLE_JEMALLOC=1 make valgrind_check; valgrind db_bench
4. make clean; COMPILE_WITH_TSAN=1 make all check; db_bench
5. make clean; COMPILE_WITH_ASAN=1 make all check; db_bench
6. make clean; OPT=-DROCKSDB_LITE make check
7. verify no perf regressions when disabled
Reviewers: igor, sdong
Reviewed By: sdong
Subscribers: MarkCallaghan, IslamAbdelRahman, anthony, yhchiang, rven, sdong, guyg8, kradhakrishnan, dhruba
Differential Revision: https://reviews.facebook.net/D50589
Summary: We now have a mechanism to further slowdown writes. Double default options.delayed_write_rate to try to keep the default behavior closer to it used to be.
Test Plan: Run all tests.
Reviewers: IslamAbdelRahman
Reviewed By: IslamAbdelRahman
Subscribers: yhchiang, kradhakrishnan, rven, leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D52281
Summary: It's usually hard for users to set a value of options.delayed_write_rate. With this diff, after slowdown condition triggers, we greedily reduce write rate if estimated pending compaction bytes increase. If estimated compaction pending bytes drop, we increase the write rate.
Test Plan:
Add a unit test
Test with db_bench setting:
TEST_TMPDIR=/dev/shm/ ./db_bench --benchmarks=fillrandom -num=10000000 --soft_pending_compaction_bytes_limit=1000000000 --hard_pending_compaction_bytes_limit=3000000000 --delayed_write_rate=100000000
and make sure without the commit, write stop will happen, but with the commit, it will not happen.
Reviewers: igor, anthony, rven, yhchiang, kradhakrishnan, IslamAbdelRahman
Reviewed By: IslamAbdelRahman
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D52131
Summary:
This patch update the Iterator API to introduce new functions that allow users to keep the Slices returned by key() valid as long as the Iterator is not deleted
ReadOptions::pin_data : If true keep loaded blocks in memory as long as the iterator is not deleted
Iterator::IsKeyPinned() : If true, this mean that the Slice returned by key() is valid as long as the iterator is not deleted
Also add a new option BlockBasedTableOptions::use_delta_encoding to allow users to disable delta_encoding if needed.
Benchmark results (using https://phabricator.fb.com/P20083553)
```
// $ du -h /home/tec/local/normal.4K.Snappy/db10077
// 6.1G /home/tec/local/normal.4K.Snappy/db10077
// $ du -h /home/tec/local/zero.8K.LZ4/db10077
// 6.4G /home/tec/local/zero.8K.LZ4/db10077
// Benchmarks for shard db10077
// _build/opt/rocks/benchmark/rocks_copy_benchmark \
// --normal_db_path="/home/tec/local/normal.4K.Snappy/db10077" \
// --zero_db_path="/home/tec/local/zero.8K.LZ4/db10077"
// First run
// ============================================================================
// rocks/benchmark/RocksCopyBenchmark.cpp relative time/iter iters/s
// ============================================================================
// BM_StringCopy 1.73s 576.97m
// BM_StringPiece 103.74% 1.67s 598.55m
// ============================================================================
// Match rate : 1000000 / 1000000
// Second run
// ============================================================================
// rocks/benchmark/RocksCopyBenchmark.cpp relative time/iter iters/s
// ============================================================================
// BM_StringCopy 611.99ms 1.63
// BM_StringPiece 203.76% 300.35ms 3.33
// ============================================================================
// Match rate : 1000000 / 1000000
```
Test Plan: Unit tests
Reviewers: sdong, igor, anthony, yhchiang, rven
Reviewed By: rven
Subscribers: dhruba, lovro, adsharma
Differential Revision: https://reviews.facebook.net/D48999
Summary: Deprecate options.soft_rate_limit, which is hard to tune, with options.soft_pending_compaction_bytes_limit, which would trigger the slowdown if estimated pending compaction bytes exceeds the threshold. The hope is to make it more striaght-forward to tune.
Test Plan: Modify DBTest.SoftLimit to cover options.soft_pending_compaction_bytes_limit instead; run all unit tests.
Reviewers: IslamAbdelRahman, yhchiang, rven, kradhakrishnan, igor, anthony
Reviewed By: anthony
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D51117
Summary: Introduce a compaction picking priority that picks files who contains the oldest rows to compact. This is a mode that slightly improves write amplification for random update cases.
Test Plan: Add a unit test and run it in valgrind too.
Reviewers: yhchiang, anthony, IslamAbdelRahman, rven, kradhakrishnan, MarkCallaghan, igor
Reviewed By: igor
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D51459
Summary: options.row_cache should already been initialized as null by default. Still try to set it following current convention, because one valgrind failure reports a failure related to it.
Test Plan: Run all unit tests
Reviewers: yhchiang, kradhakrishnan, igor
Reviewed By: igor
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D51303
Summary:
This patch allows rocksdb to persist options into a file on
DB::Open, SetOptions, and Create / Drop ColumnFamily.
Options files are created under the same directory as the rocksdb
instance.
In addition, this patch also adds a fail_if_missing_options_file in DBOptions
that makes any function call return non-ok status when it is not able to
persist options properly.
// If true, then DB::Open / CreateColumnFamily / DropColumnFamily
// / SetOptions will fail if options file is not detected or properly
// persisted.
//
// DEFAULT: false
bool fail_if_missing_options_file;
Options file names are formatted as OPTIONS-<number>, and RocksDB
will always keep the latest two options files.
Test Plan:
Add options_file_test.
options_test
column_family_test
Reviewers: igor, IslamAbdelRahman, sdong, anthony
Reviewed By: anthony
Subscribers: dhruba
Differential Revision: https://reviews.facebook.net/D48285
Enable C4307 'operator' : integral constant overflow
Longs and ints on Windows are 32-bit hence the overflow
Enable C4309 'conversion' : truncation of constant value
Enable C4512 'class' : assignment operator could not be generated
Enable C4701 Potentially uninitialized local variable 'name' used
Summary:
MyRocks testing found an issue that while iterating over keys
that are outside the prefix, sometimes wrong results were seen for keys
outside the prefix. We now tighten the range of keys seen with a new
read option called prefix_seen_at_start. This remembers the starting
prefix and then compares it on a Next for equality of prefix. If they
are from a different prefix, it sets valid to false.
Test Plan: PrefixTest.PrefixValid
Reviewers: IslamAbdelRahman, sdong, yhchiang, anthony
Reviewed By: anthony
Subscribers: spetrunia, hermanlee4, yoshinorim, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D50211
Summary: Run "make format" for some recent commits.
Test Plan: Build and run tests
Reviewers: IslamAbdelRahman
Reviewed By: IslamAbdelRahman
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D49707
introduce a new DBOption random_access_max_buffer_size to limit
the size of the random access buffer used for unbuffered access.
Implement read ahead buffering when enabled.
To that effect propagate compaction_readahead_size and the new option
to the env options to make it available for the implementation.
Add Hint() override so SetupForCompaction() call would call Hint()
readahead can now be setup from both Hint() and EnableReadAhead()
Add new option random_access_max_buffer_size support
db_bench, options_helper to make it string parsable
and the unit test.
Summary:
Added boolean variable to guard fallocate() calls.
Set to false to prevent space leaks when tests fail.
Test Plan:
Compliles
Set to false and ran log device tests
Reviewers: sdong, lovro, igor
Reviewed By: igor
Subscribers: dhruba
Differential Revision: https://reviews.facebook.net/D48027
Summary:
Add options.compaction_pri, which specifies the policy about which file to compact first.
kCompactionPriByLargestSeq will compact oldest files first.
Verified the behavior in db_bench but did not write unit tests yet. Also need to make it settable through option string and dynamically changeable.
Test Plan: Will write unit tests
Reviewers: igor, rven, anthony, kradhakrishnan, IslamAbdelRahman, yhchiang, MarkCallaghan
Reviewed By: yhchiang
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D45951
Summary: Change the log level of DB start-up log from Warn to Header.
Test Plan: db_bench and observe the LOG header
Reviewers: igor, anthony, IslamAbdelRahman, sdong
Reviewed By: sdong
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D47067
Summary:
RocksDB options can be dumped to the log file, and
up to this point the max_subcompactions option was not included
in this dump. This fixes that.
Test Plan: makek all && make check
Reviewers: MarkCallaghan, igor, noetzli, anthony, yhchiang, sdong
Reviewed By: yhchiang, sdong
Subscribers: dhruba
Differential Revision: https://reviews.facebook.net/D46971
Summary: Add an option to stop writes if compaction lefts behind. If estimated pending compaction bytes is more than threshold specified by options.hard_pending_compaction_bytes_liimt, writes will stop until compactions are cleared to under the threshold.
Test Plan: Add unit test DBTest.HardLimit
Reviewers: rven, kradhakrishnan, anthony, IslamAbdelRahman, yhchiang, igor
Reviewed By: igor
Subscribers: MarkCallaghan, leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D45999
Summary:
ReadaheadRandomAccessFile acts as a transparent layer on top of RandomAccessFile. When a Read() request is issued, it issues a much bigger request to the OS and caches the result. When a new request comes in and we already have the data cached, it doesn't have to issue any requests to the OS.
We add ReadaheadRandomAccessFile layer only when file is read during compactions.
D45105 was incorrectly closed by Phabricator because I committed it to a separate branch (not master), so I'm resubmitting the diff.
Test Plan: make check
Reviewers: MarkCallaghan, sdong
Reviewed By: sdong
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D45123
Summary:
Up until this point we had DbOptions.num_subcompactions, but
it is semantically more correct to call this max_subcompactions since
we will schedule *up to* DbOptions.max_subcompactions smaller compactions
at a time during a compaction job.
I also added a --subcompactions option to db_bench
Test Plan: make all make check
Reviewers: sdong, igor, anthony, yhchiang
Reviewed By: yhchiang
Subscribers: dhruba
Differential Revision: https://reviews.facebook.net/D45069
Summary: Currently compaction inputs share the same file descriptor and table reader as other foreground threads. It makes fadvise works less predictable. Add options.new_table_reader_for_compaction_inputs to enforce to create a new file descriptor and new table reader for it.
Test Plan: Add the option.
Reviewers: rven, anthony, kradhakrishnan, IslamAbdelRahman, igor, yhchiang
Reviewed By: igor
Subscribers: igor, MarkCallaghan, leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D43311
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: 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:
wal_recovery_mode setting was not written to LOG. This diff
adds the log message
Test Plan: manually checked
Reviewers: kradhakrishnan, sdong, igor
Reviewed By: igor
Subscribers: dhruba
Differential Revision: https://reviews.facebook.net/D43953
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:
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:
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:
When we first started, max_background_flushes was 0 by default and compaction thread was executing flushes (since there was no flush thread). Then, we switched the default max_background_flushes to 1. However, we still support the case where there is no flush thread and flushes are done in compaction. This is making our code a bit more complicated. By not supporting this use-case we can make our code simpler.
We have a special case that when you set max_background_flushes to 0, we
schedule the flush to execute on the compaction thread.
Test Plan: make check (there might be some unit tests that depend on this behavior)
Reviewers: IslamAbdelRahman, yhchiang, sdong
Reviewed By: sdong
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D41931
Summary: It has been around for a while and it looks like it never found any uses in the wild. It's also complicating our compaction_job code quite a bit. We're deprecating it in 3.13, but will put it back in 3.14 if we actually find users that need this feature.
Test Plan: make check
Reviewers: noetzli, yhchiang, sdong
Reviewed By: sdong
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D42405
Summary: This option is guarding the feature implemented 2 and a half years ago: D8991. The feature was enabled by default back then and has been running without issues. There is no reason why any client would turn this feature off. I found no reference in fbcode.
Test Plan: none
Reviewers: sdong, yhchiang, anthony, dhruba
Reviewed By: dhruba
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D42063
Summary: This helps Windows port to format their changes, as discussed. Might have formatted some other codes too becasue last 10 commits include more.
Test Plan: Build it.
Reviewers: anthony, IslamAbdelRahman, kradhakrishnan, yhchiang, igor
Reviewed By: igor
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D41961
invalid suffix on literal
no return statement in function returning non-void CuckooStep::operator=
extra qualification ‘rocksdb::spatial::Variant::
dereferencing type-punned pointer will break strict-aliasing rules
Summary: Make RocksDb build and run on Windows to be functionally
complete and performant. All existing test cases run with no
regressions. Performance numbers are in the pull-request.
Test plan: make all of the existing unit tests pass, obtain perf numbers.
Co-authored-by: Praveen Rao praveensinghrao@outlook.com
Co-authored-by: Sherlock Huang baihan.huang@gmail.com
Co-authored-by: Alex Zinoviev alexander.zinoviev@me.com
Co-authored-by: Dmitri Smirnov dmitrism@microsoft.com
Summary:
Implementation of a table-level row cache.
It only caches point queries done through the `DB::Get` interface, queries done through the `Iterator` interface will completely skip the cache.
Supports snapshots and merge operations.
Test Plan: Ran `make valgrind_check commit-prereq`
Reviewers: igor, philipp, sdong
Reviewed By: sdong
Subscribers: dhruba
Differential Revision: https://reviews.facebook.net/D39849
Summary:
The "one size fits all" approach with WAL recovery will only introduce inconvenience for our varied clients as we go forward. The current recovery is a bit heuristic. We introduce the following levels of consistency while replaying the WAL.
1. RecoverAfterRestart (kTolerateCorruptedTailRecords)
This mocks the current recovery mode.
2. RecoverAfterCleanShutdown (kAbsoluteConsistency)
This is ideal for unit test and cases where the store is shutdown cleanly. We tolerate no corruption or incomplete writes.
3. RecoverPointInTime (kPointInTimeRecovery)
This is ideal when using devices with controller cache or file systems which can loose data on restart. We recover upto the point were is no corruption or incomplete write.
4. RecoverAfterDisaster (kSkipAnyCorruptRecord)
This is ideal mode to recover data. We tolerate corruption and incomplete writes, and we hop over those sections that we cannot make sense of salvaging as many records as possible.
Test Plan:
(1) Run added unit test to cover all levels.
(2) Run make check.
Reviewers: leveldb, sdong, igor
Subscribers: yoshinorim, dhruba
Differential Revision: https://reviews.facebook.net/D38487
Summary:
Currently RocksDB silently ignores this issue and doesn't compress the data. Based on discussion, we agree that this is pretty bad because it can cause confusion for our users.
This patch fails DB::Open() if we don't support the compression that is specified in the options.
Test Plan: make check with LZ4 not present. If Snappy is not present all tests will just fail because Snappy is our default library. We should make Snappy the requirement, since without it our default DB::Open() fails.
Reviewers: sdong, MarkCallaghan, rven, yhchiang
Reviewed By: yhchiang
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D39687
Summary: Currently we dump DBOptions for each column family options we dump. This leads to duplicate lines in our LOG file. This diff fixes that.
Test Plan: Check out the LOG
Reviewers: sdong, rven, yhchiang
Reviewed By: yhchiang
Subscribers: IslamAbdelRahman, yoshinorim, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D39729
Summary:
We slow down data into the database to the rate of options.delayed_write_rate (a new option) with this patch.
The thread synchronization approach I take is to still synchronize write controller by DB mutex and GetDelay() is inside DB mutex. Try to minimize the frequency of getting time in GetDelay(). I verified it through db_bench and it seems to work
hard_rate_limit is deprecated.
options.delayed_write_rate is still not dynamically changeable. Need to work on it as a follow-up.
Test Plan: Add new unit tests in db_test
Reviewers: yhchiang, rven, kradhakrishnan, anthony, MarkCallaghan, igor
Reviewed By: igor
Subscribers: ikabiljo, leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D36351
Summary:
Replacing the default value for compaction_filter_factory and compaction_filter_factory_v2 to be nullptr instead of DefaultCompactionFilterFactory / DefaultCompactionFilterFactoryV2
The reason for this is to be able to determine easily if we have compaction filter factory or not without depending on RTTI
Test Plan: make check
Reviewers: yoshinorim, ott, igor, sdong
Reviewed By: sdong
Subscribers: dhruba
Differential Revision: https://reviews.facebook.net/D39693
