Summary:
Context:
As need for new feature of resource management using RocksDB's rate limiter like [https://github.com/facebook/rocksdb/issues/8595](https://github.com/facebook/rocksdb/pull/8595) arises, it is about time to re-learn our rate limiter and make this learning process easier for others by improving its readability. The comment/assertion/one extra else-branch are added based on my best understanding toward the rate_limiter.cc and rate_limiter_test.cc up to date after giving it a hard read.
- Add code comments/assertion/one extra else-branch (that is not affecting existing behavior, see PR comment) to describe how leader-election works under multi-thread settings in GenericRateLimiter::Request()
- Add code comments to describe a non-obvious trick during clean-up of rate limiter destructor
- Add code comments to explain more about the starvation being fixed in GenericRateLimiter::Refill() through partial byte-granting
- Add code comments to the rate limiter's setup in a complicated unit test in rate_limiter_test
Pull Request resolved: https://github.com/facebook/rocksdb/pull/8596
Test Plan: - passed existing rate_limiter_test.cc
Reviewed By: ajkr
Differential Revision: D29982590
Pulled By: hx235
fbshipit-source-id: c3592986bb5b0c90d8229fe44f425251ec7e8a0a
Summary:
Calling the GetImpl function could leave reference to a local
callback function in a field of a parameter struct. As this is
performance-critical code, I'm not going to attempt to sanitize this
code too much, but make the existing hack a bit cleaner by reverting
what it overwrites in the input struct.
Added SaveAndRestore utility class to make that easier.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/8590
Test Plan:
added unit test for SaveAndRestore; existing tests for
GetImpl
Reviewed By: riversand963
Differential Revision: D29947983
Pulled By: pdillinger
fbshipit-source-id: 2f608853f970bc06724e834cc84dcc4b8599ddeb
Summary:
The PerThreadDBPath has already specified a slash. It does not need to be specified when initializing the test path.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/8555
Reviewed By: ajkr
Differential Revision: D29758399
Pulled By: jay-zhuang
fbshipit-source-id: 6d2b878523e3e8580536e2829cb25489844d9011
Summary:
Rare TSAN and valgrind failures are caused by unnecessary
reading of a field on the TaskLimiterToken::limiter_ for an assertion
after the token has been released and the limiter destroyed. To simplify
we can simply destroy the token before triggering DB shutdown
(potentially destroying the limiter). This makes the ReleaseOnce logic
unnecessary.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/8567
Test Plan: watch for more failures in CI
Reviewed By: ajkr
Differential Revision: D29811795
Pulled By: pdillinger
fbshipit-source-id: 135549ebb98fe4f176d1542ed85d5bd6350a40b3
Summary:
Fixed a few MSVC (VCToolsVersion=14.0) build errors and warnings
* `DEFINE_string` is a macro and VC compiler complains that it cannot put [ifdef-inside-define](https://stackoverflow.com/questions/5586429/ifdef-inside-define)
* `sleep()` is not a recognizable function. Use `FLAGS_env->SleepForMicroseconds` instead
* Define precise type in comparison to avoid mismatch warning
Pull Request resolved: https://github.com/facebook/rocksdb/pull/8519
Reviewed By: jay-zhuang
Differential Revision: D29683086
fbshipit-source-id: 8c80941472089f8daba84ae29597e75e603850e4
Summary:
In PR https://github.com/facebook/rocksdb/issues/7523 , checksum handoff is introduced in RocksDB for WAL, Manifest, and SST files. When user enable checksum handoff for a certain type of file, before the data is written to the lower layer storage system, we calculate the checksum (crc32c) of each piece of data and pass the checksum down with the data, such that data verification can be down by the lower layer storage system if it has the capability. However, it cannot cover the whole lifetime of the data in the memory and also it potentially introduces extra checksum calculation overhead.
In this PR, we introduce a new interface in WritableFileWriter::Append, which allows the caller be able to pass the data and the checksum (crc32c) together. In this way, WritableFileWriter can directly use the pass-in checksum (crc32c) to generate the checksum of data being passed down to the storage system. It saves the calculation overhead and achieves higher protection coverage. When a new checksum is added with the data, we use Crc32cCombine https://github.com/facebook/rocksdb/issues/8305 to combine the existing checksum and the new checksum. To avoid the segmenting of data by rate-limiter before it is stored, rate-limiter is called enough times to accumulate enough credits for a certain write. This design only support Manifest and WAL which use log_writer in the current stage.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/8412
Test Plan: make check, add new testing cases.
Reviewed By: anand1976
Differential Revision: D29151545
Pulled By: zhichao-cao
fbshipit-source-id: 75e2278c5126cfd58393c67b1efd18dcc7a30772
Summary:
Implement a function to generate the crc32c of two combined strings. Suppose we have the string 1 (s1) with crc32c checksum crc32c_1 and string 2 (s2) with crc32c checksum crc32c_2, the new string is s1+s2 and its checksum is crc32c_new=Crc32cCombine(crc32c_1, crc32c_2, s2.size).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/8305
Test Plan: make check, added new testing case
Reviewed By: pdillinger
Differential Revision: D28651665
Pulled By: zhichao-cao
fbshipit-source-id: c84116108388f11a81f6a217b49f99c70d4ffacf
Summary:
Makes the Comparator class into a Customizable object. Added/Updated the CreateFromString method to create Comparators. Added test for using the ObjectRegistry to create one.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/8336
Reviewed By: jay-zhuang
Differential Revision: D28999612
Pulled By: mrambacher
fbshipit-source-id: bff2cb2814eeb9fef6a00fddc61d6e34b6fbcf2e
Summary:
- Add class `FunctorWrapper` to invoke the function with given parameters
- Implement `StartThreadTyped` which wraps `StartThread` with type checking cover
- Demonstrate `StartThreadTyped` in test `util/thread_local_test.cc`
https://github.com/facebook/rocksdb/issues/8285
Pull Request resolved: https://github.com/facebook/rocksdb/pull/8303
Reviewed By: ajkr
Differential Revision: D28539318
Pulled By: pdillinger
fbshipit-source-id: 624789c236bde31163deda95c1e1471aee68933e
Summary:
The functions will be used for remote compaction parameter
input and result.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/8247
Test Plan: `make check`
Reviewed By: ajkr
Differential Revision: D28104680
Pulled By: jay-zhuang
fbshipit-source-id: c0a5178e6277125118384278efea2acbf90aa6cb
Summary:
Previously the shutdown process did not properly wait for all
`compaction_thread_limiter` tokens to be released before proceeding to
delete the DB's C++ objects. When this happened, we saw tests like
"DBCompactionTest.CompactionLimiter" flake with the following error:
```
virtual
rocksdb::ConcurrentTaskLimiterImpl::~ConcurrentTaskLimiterImpl():
Assertion `outstanding_tasks_ == 0' failed.
```
There is a case where a token can still be alive even after the shutdown
process has waited for BG work to complete. In particular, this happens
because the shutdown process only waits for flush/compaction scheduled/unscheduled counters to all
reach zero. These counters are decremented in `BackgroundCallCompaction()`
functions. However, tokens are released in `BGWork*Compaction()` functions, which
actually wrap the `BackgroundCallCompaction()` function.
A simple sleep could repro the race condition:
```
$ diff --git a/db/db_impl/db_impl_compaction_flush.cc
b/db/db_impl/db_impl_compaction_flush.cc
index 806bc548a..ba59efa89 100644
--- a/db/db_impl/db_impl_compaction_flush.cc
+++ b/db/db_impl/db_impl_compaction_flush.cc
@@ -2442,6 +2442,7 @@ void DBImpl::BGWorkCompaction(void* arg) {
static_cast<PrepickedCompaction*>(ca.prepicked_compaction);
static_cast_with_check<DBImpl>(ca.db)->BackgroundCallCompaction(
prepicked_compaction, Env::Priority::LOW);
+ sleep(1);
delete prepicked_compaction;
}
$ ./db_compaction_test --gtest_filter=DBCompactionTest.CompactionLimiter
db_compaction_test: util/concurrent_task_limiter_impl.cc:24: virtual rocksdb::ConcurrentTaskLimiterImpl::~ConcurrentTaskLimiterImpl(): Assertion `outstanding_tasks_ == 0' failed.
Received signal 6 (Aborted)
#0 /usr/local/fbcode/platform007/lib/libc.so.6(gsignal+0xcf) [0x7f02673c30ff] ?? ??:0
https://github.com/facebook/rocksdb/issues/1 /usr/local/fbcode/platform007/lib/libc.so.6(abort+0x134) [0x7f02673ac934] ?? ??:0
...
```
Pull Request resolved: https://github.com/facebook/rocksdb/pull/8253
Test Plan: sleeps to expose race conditions
Reviewed By: akankshamahajan15
Differential Revision: D28168064
Pulled By: ajkr
fbshipit-source-id: 9e5167c74398d323e7975980c5cc00f450631160
Summary:
Renaming ImmutableCFOptions::info_log and statistics to logger and stats. This is stage 2 in creating an ImmutableOptions class. It is necessary because the names match those in ImmutableOptions and have different types.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/8227
Reviewed By: jay-zhuang
Differential Revision: D28000967
Pulled By: mrambacher
fbshipit-source-id: 3bf2aa04e8f1e8724d825b7deacf41080c14420b
Summary:
This PR is a first step at attempting to clean up some of the Mutable/Immutable Options code. With this change, a DBOption and a ColumnFamilyOption can be reconstructed from their Mutable and Immutable equivalents, respectively.
readrandom tests do not show any performance degradation versus master (though both are slightly slower than the current 6.19 release).
There are still fields in the ImmutableCFOptions that are not CF options but DB options. Eventually, I would like to move those into an ImmutableOptions (= ImmutableDBOptions+ImmutableCFOptions). But that will be part of a future PR to minimize changes and disruptions.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/8176
Reviewed By: pdillinger
Differential Revision: D27954339
Pulled By: mrambacher
fbshipit-source-id: ec6b805ba9afe6e094bffdbd76246c2d99aa9fad
Summary:
This partially reverts commit 10196d7edc.
The problem with this change is because of important filter use cases:
FIFO compaction and SST writer. FIFO "compaction" always uses level 0 so
would only use Ribbon filters if specifically including level 0 for the
Ribbon filter policy. SST writer sets level_at_creation=-1 to indicate
unknown level, and this would be treated the same as level 0 unless
fixed.
We are keeping the part about committing to permanent schema, which is
only changes to API comments and HISTORY.md.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/8212
Test Plan: CI
Reviewed By: jay-zhuang
Differential Revision: D27896468
Pulled By: pdillinger
fbshipit-source-id: 50a775f7cba5d64fb729d9b982e355864020596e
Summary:
Since the Ribbon filter schema seems good (compatible back to
6.15.0), this change commits to long term support of the SST schema,
even though we expect the API for enabling Ribbon to change (still
called NewExperimentalRibbonFilterPolicy).
This also adds support for "hybrid" configuration in which some levels
use Bloom (higher levels, lower numbered) for speed and the rest use
Ribbon (lower levels, higher numbered) for memory space efficiency.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/8198
Test Plan: unit test added, crash test support
Reviewed By: jay-zhuang
Differential Revision: D27831232
Pulled By: pdillinger
fbshipit-source-id: 90e528677689474d293ed6710b42ba89fbd5b5ab
Summary:
The patch adds a resource management/RAII class called `ThreadGuard`,
which can be used to ensure that the managed thread is joined when the
`ThreadGuard` is destroyed, regardless of whether it is due to the
object going out of scope, an early return, an exception etc. This is
important because if an `std::thread` object is destroyed without having
been joined (or detached) first, the process is aborted (via
`std::terminate`).
For now, `ThreadGuard` is only used in the test case
`ExternalSSTFileTest.PickedLevelBug`; however, it could come in handy
elsewhere in the codebase as well (both in test code and "real" code).
Case in point: in the `PickedLevelBug` test case, with the earlier code we
could end up in the above situation when the following assertion (which is
before the threads are joined) is triggered:
```
ASSERT_FALSE(bg_compact_started.load());
```
Pull Request resolved: https://github.com/facebook/rocksdb/pull/8112
Test Plan:
```
make check
gtest-parallel --repeat=10000 ./external_sst_file_test --gtest_filter="*PickedLevelBug"
```
Reviewed By: riversand963
Differential Revision: D27343185
Pulled By: ltamasi
fbshipit-source-id: 2a8c3aa68bc78cc03ec0dbae909fb25c2cd15c69
Summary:
`strerror()` is not thread-safe, using `strerror_r()` instead. The API could be different on the different platforms, used the code from 0deef031cb/folly/String.cpp (L457)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/8087
Reviewed By: mrambacher
Differential Revision: D27267151
Pulled By: jay-zhuang
fbshipit-source-id: 4b8856d1ec069d5f239b764750682c56e5be9ddb
Summary:
Improved handling of -bits_per_key other than 10, but at least
the OptimizeForMemory test is simply not designed for generally handling
other settings. (ribbon_test does have a statistical framework for this
kind of testing, but it's not important to do that same for Bloom right
now.)
Closes https://github.com/facebook/rocksdb/issues/7019
Pull Request resolved: https://github.com/facebook/rocksdb/pull/8093
Test Plan: for I in `seq 1 20`; do ./bloom_test --gtest_filter=-*OptimizeForMemory* --bits_per_key=$I &> /dev/null || echo FAILED; done
Reviewed By: mrambacher
Differential Revision: D27275875
Pulled By: pdillinger
fbshipit-source-id: 7362e8ac2c41ea11f639412e4f30c8b375f04388
Summary:
For performance purposes, the lower level routines were changed to use a SystemClock* instead of a std::shared_ptr<SystemClock>. The shared ptr has some performance degradation on certain hardware classes.
For most of the system, there is no risk of the pointer being deleted/invalid because the shared_ptr will be stored elsewhere. For example, the ImmutableDBOptions stores the Env which has a std::shared_ptr<SystemClock> in it. The SystemClock* within the ImmutableDBOptions is essentially a "short cut" to gain access to this constant resource.
There were a few classes (PeriodicWorkScheduler?) where the "short cut" property did not hold. In those cases, the shared pointer was preserved.
Using db_bench readrandom perf_level=3 on my EC2 box, this change performed as well or better than 6.17:
6.17: readrandom : 28.046 micros/op 854902 ops/sec; 61.3 MB/s (355999 of 355999 found)
6.18: readrandom : 32.615 micros/op 735306 ops/sec; 52.7 MB/s (290999 of 290999 found)
PR: readrandom : 27.500 micros/op 871909 ops/sec; 62.5 MB/s (367999 of 367999 found)
(Note that the times for 6.18 are prior to revert of the SystemClock).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/8033
Reviewed By: pdillinger
Differential Revision: D27014563
Pulled By: mrambacher
fbshipit-source-id: ad0459eba03182e454391b5926bf5cdd45657b67
Summary:
This is for cases that do not meet the Facebook criteria for
SKIP (see new comments). Also made ROCKSDB_GTEST_{SKIP,BYPASS} print the
message because gtest doesn't ever seem to.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/8048
Test Plan: manual inspection of ./ribbon_test output, CI
Reviewed By: mrambacher
Differential Revision: D26953688
Pulled By: pdillinger
fbshipit-source-id: c914eaffe7d419db6ab90a193d474531e23582e5
Summary:
Removed confusing, awkward, and undocumented internal API
ReadOneLine and replaced with very simple LineFileReader.
In refactoring backupable_db.cc, this has the side benefit of
removing the arbitrary cap on the size of backup metadata files.
Also added Status::MustCheck to make it easy to mark a Status as
"must check." Using this, I can ensure that after
LineFileReader::ReadLine returns false the caller checks GetStatus().
Also removed some excessive conditional compilation in status.h
Pull Request resolved: https://github.com/facebook/rocksdb/pull/8026
Test Plan: added unit test, and running tests with ASSERT_STATUS_CHECKED
Reviewed By: mrambacher
Differential Revision: D26831687
Pulled By: pdillinger
fbshipit-source-id: ef749c265a7a26bb13cd44f6f0f97db2955f6f0f
Summary:
The patch does the following:
1) Exposes the amount of data (number of bytes) read from blob files from
`BlobFileReader::GetBlob` / `Version::GetBlob`.
2) Tracks the total number and size of blobs read from blob files during a
compaction (due to garbage collection or compaction filter usage) in
`CompactionIterationStats` and propagates this data to
`InternalStats::CompactionStats` / `CompactionJobStats`.
3) Updates the formulae for write amplification calculations to include the
amount of data read from blob files.
4) Extends the compaction stats dump with a new column `Rblob(GB)` and
a new line containing the total number and size of blob files in the current
`Version` to complement the information about the shape and size of the LSM tree
that's already there.
5) Updates `CompactionJobStats` so that the number of files and amount of data
written by a compaction are broken down per file type (i.e. table/blob file).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/8022
Test Plan: Ran `make check` and `db_bench`.
Reviewed By: riversand963
Differential Revision: D26801199
Pulled By: ltamasi
fbshipit-source-id: 28a5f072048a702643b28cb5971b4099acabbfb2
Summary:
This change only affects non-schema-critical aspects of the production candidate Ribbon filter. Specifically, it refines choice of internal configuration parameters based on inputs. The changes are minor enough that the schema tests in bloom_test, some of which depend on this, are unaffected. There are also some minor optimizations and refactorings.
This would be a schema change for "smash" Ribbon, to fix some known issues with small filters, but "smash" Ribbon is not accessible in public APIs. Unit test CompactnessAndBacktrackAndFpRate updated to test small and medium-large filters. Run with --thoroughness=100 or so for much better detection power (not appropriate for continuous regression testing).
Homogenous Ribbon:
This change adds internally a Ribbon filter variant we call Homogeneous Ribbon, in collaboration with Stefan Walzer. The expected "result" value for every key is zero, instead of computed from a hash. Entropy for queries not to be false positives comes from free variables ("overhead") in the solution structure, which are populated pseudorandomly. Construction is slightly faster for not tracking result values, and never fails. Instead, FP rate can jump up whenever and whereever entries are packed too tightly. For small structures, we can choose overhead to make this FP rate jump unlikely, as seen in updated unit test CompactnessAndBacktrackAndFpRate.
Unlike standard Ribbon, Homogeneous Ribbon seems to scale to arbitrary number of keys when accepting an FP rate penalty for small pockets of high FP rate in the structure. For example, 64-bit ribbon with 8 solution columns and 10% allocated space overhead for slots seems to achieve about 10.5% space overhead vs. information-theoretic minimum based on its observed FP rate with expected pockets of degradation. (FP rate is close to 1/256.) If targeting a higher FP rate with fewer solution columns, Homogeneous Ribbon can be even more space efficient, because the penalty from degradation is relatively smaller. If targeting a lower FP rate, Homogeneous Ribbon is less space efficient, as more allocated overhead is needed to keep the FP rate impact of degradation relatively under control. The new OptimizeHomogAtScale tool in ribbon_test helps to find these optimal allocation overheads for different numbers of solution columns. And Ribbon widths, with 128-bit Ribbon apparently cutting space overheads in half vs. 64-bit.
Other misc item specifics:
* Ribbon APIs in util/ribbon_config.h now provide configuration data for not just 5% construction failure rate (95% success), but also 50% and 0.1%.
* Note that the Ribbon structure does not exhibit "threshold" behavior as standard Xor filter does, so there is a roughly fixed space penalty to cut construction failure rate in half. Thus, there isn't really an "almost sure" setting.
* Although we can extrapolate settings for large filters, we don't have a good formula for configuring smaller filters (< 2^17 slots or so), and efforts to summarize with a formula have failed. Thus, small data is hard-coded from updated FindOccupancy tool.
* Enhances ApproximateNumEntries for public API Ribbon using more precise data (new API GetNumToAdd), thus a more accurate but not perfect reversal of CalculateSpace. (bloom_test updated to expect the greater precision)
* Move EndianSwapValue from coding.h to coding_lean.h to keep Ribbon code easily transferable from RocksDB
* Add some missing 'const' to member functions
* Small optimization to 128-bit BitParity
* Small refactoring of BandingStorage in ribbon_alg.h to support Homogeneous Ribbon
* CompactnessAndBacktrackAndFpRate now has an "expand" test: on construction failure, a possible alternative to re-seeding hash functions is simply to increase the number of slots (allocated space overhead) and try again with essentially the same hash values. (Start locations will be different roundings of the same scaled hash values--because fastrange not mod.) This seems to be as effective or more effective than re-seeding, as long as we increase the number of slots (m) by roughly m += m/w where w is the Ribbon width. This way, there is effectively an expansion by one slot for each ribbon-width window in the banding. (This approach assumes that getting "bad data" from your hash function is as unlikely as it naturally should be, e.g. no adversary.)
* 32-bit and 16-bit Ribbon configurations are added to ribbon_test for understanding their behavior, e.g. with FindOccupancy. They are not considered useful at this time and not tested with CompactnessAndBacktrackAndFpRate.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7879
Test Plan: unit test updates included
Reviewed By: jay-zhuang
Differential Revision: D26371245
Pulled By: pdillinger
fbshipit-source-id: da6600d90a3785b99ad17a88b2a3027710b4ea3a
Summary:
Fixed 5 test case failures found on Windows 10/Windows Server 2016
1. In `flush_job_test`, the DestroyDir function fails in deconstructor because some file handles are still being held by VersionSet. This happens on Windows Server 2016, so need to manually reset versions_ pointer to release all file handles.
2. In `StatsHistoryTest.InMemoryStatsHistoryPurging` test, the capping memory cost of stats_history_size on Windows becomes 14000 bytes with latest changes, not just 13000 bytes.
3. In `SSTDumpToolTest.RawOutput` test, the output file handle is not closed at the end.
4. In `FullBloomTest.OptimizeForMemory` test, ROCKSDB_MALLOC_USABLE_SIZE is undefined on windows so `total_mem` is always equal to `total_size`. The internal memory fragmentation assertion does not apply in this case.
5. In `BlockFetcherTest.FetchAndUncompressCompressedDataBlock` test, XPRESS cannot reach 87.5% compression ratio with original CreateTable method, so I append extra zeros to the string value to enhance compression ratio. Beside, since XPRESS allocates memory internally, thus does not support for custom allocator verification, we will skip the allocator verification for XPRESS
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7992
Reviewed By: jay-zhuang
Differential Revision: D26615283
Pulled By: ajkr
fbshipit-source-id: 3632612f84b99e2b9c77c403b112b6bedf3b125d
Summary:
For dictionary compression, we need to collect some representative samples of the data to be compressed, which we use to either generate or train (when `CompressionOptions::zstd_max_train_bytes > 0`) a dictionary. Previously, the strategy was to buffer all the data blocks during flush, and up to the target file size during compaction. That strategy allowed us to randomly pick samples from as wide a range as possible that'd be guaranteed to land in a single output file.
However, some users try to make huge files in memory-constrained environments, where this strategy can cause OOM. This PR introduces an option, `CompressionOptions::max_dict_buffer_bytes`, that limits how much data blocks are buffered before we switch to unbuffered mode (which means creating the per-SST dictionary, writing out the buffered data, and compressing/writing new blocks as soon as they are built). It is not strict as we currently buffer more than just data blocks -- also keys are buffered. But it does make a step towards giving users predictable memory usage.
Related changes include:
- Changed sampling for dictionary compression to select unique data blocks when there is limited availability of data blocks
- Made use of `BlockBuilder::SwapAndReset()` to save an allocation+memcpy when buffering data blocks for building a dictionary
- Changed `ParseBoolean()` to accept an input containing characters after the boolean. This is necessary since, with this PR, a value for `CompressionOptions::enabled` is no longer necessarily the final component in the `CompressionOptions` string.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7970
Test Plan:
- updated `CompressionOptions` unit tests to verify limit is respected (to the extent expected in the current implementation) in various scenarios of flush/compaction to bottommost/non-bottommost level
- looked at jemalloc heap profiles right before and after switching to unbuffered mode during flush/compaction. Verified memory usage in buffering is proportional to the limit set.
Reviewed By: pdillinger
Differential Revision: D26467994
Pulled By: ajkr
fbshipit-source-id: 3da4ef9fba59974e4ef40e40c01611002c861465
Summary:
The current implementation of a binary heap in `util/heap.h` does a move-assign in the `pop` method. In the case that there is exactly one element stored in the heap, this ends up being a self-move-assign. This can cause trouble with certain classes, which are not prepared for this. Furthermore, it trips up the glibc STL debugger (`-D_GLIBCXX_DEBUG`), which produces an assertion failure in this case.
This PR addresses this problem by not doing the (unnecessary in this case) move-assign if there is only one element in the heap.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7942
Reviewed By: jay-zhuang
Differential Revision: D26528739
Pulled By: ajkr
fbshipit-source-id: 5ca570e0c4168f086b10308ad766dff84e6e2d03
Summary:
This will fix a missing string separation between `msg[n]` and `state_`.
Example of an error message how its looking now:
```
IO error: No space left on deviceWhile appending to file: /home/willi/src/stable-3.7/tmp/arangosh_CL6EFQ/shell_client/single1/data/engine-rocksdb/126426.sst: No space left on device
```
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7919
Reviewed By: ajkr
Differential Revision: D26242246
Pulled By: jay-zhuang
fbshipit-source-id: 5d9a0997a410aecfb3781478e57395d3d937bb84
Summary:
in PR https://github.com/facebook/rocksdb/issues/7419 , we introduce the new Append and PositionedAppend APIs to WritableFile at File System, which enable RocksDB to pass the data verification information (e.g., checksum of the data) to the lower layer. In this PR, we use the new API in WritableFileWriter, such that the file created via WritableFileWrite can pass the checksum to the storage layer. To control which types file should apply the checksum handoff, we add checksum_handoff_file_types to DBOptions. User can use this option to control which file types (Currently supported file tyes: kLogFile, kTableFile, kDescriptorFile.) should use the new Append and PositionedAppend APIs to handoff the verification information.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7523
Test Plan: add new unit test, pass make check/ make asan_check
Reviewed By: pdillinger
Differential Revision: D24313271
Pulled By: zhichao-cao
fbshipit-source-id: aafd69091ae85c3318e3e17cbb96fe7338da11d0
Summary:
Adds support for prefetching data in Ribbon queries,
which especially optimizes batched Ribbon queries for MultiGet
(~222ns/key to ~97ns/key) but also single key queries on cold memory
(~333ns to ~226ns) because many queries span more than one cache line.
This required some refactoring of the query algorithm, and there
does not appear to be a noticeable regression in "hot memory" query
times (perhaps from 48ns to 50ns).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7889
Test Plan:
existing unit tests, plus performance validation with
filter_bench:
Each data point is the best of two runs. I saturated the machine
CPUs with other filter_bench runs in the background.
Before:
$ ./filter_bench -impl=3 -m_keys_total_max=200 -average_keys_per_filter=100000 -m_queries=50
WARNING: Assertions are enabled; benchmarks unnecessarily slow
Building...
Build avg ns/key: 125.86
Number of filters: 1993
Total size (MB): 168.166
Reported total allocated memory (MB): 183.211
Reported internal fragmentation: 8.94626%
Bits/key stored: 7.05341
Prelim FP rate %: 0.951827
----------------------------
Mixed inside/outside queries...
Single filter net ns/op: 48.0111
Batched, prepared net ns/op: 222.384
Batched, unprepared net ns/op: 343.908
Skewed 50% in 1% net ns/op: 252.916
Skewed 80% in 20% net ns/op: 320.579
Random filter net ns/op: 332.957
After:
$ ./filter_bench -impl=3 -m_keys_total_max=200 -average_keys_per_filter=100000 -m_queries=50
WARNING: Assertions are enabled; benchmarks unnecessarily slow
Building...
Build avg ns/key: 128.117
Number of filters: 1993
Total size (MB): 168.166
Reported total allocated memory (MB): 183.211
Reported internal fragmentation: 8.94626%
Bits/key stored: 7.05341
Prelim FP rate %: 0.951827
----------------------------
Mixed inside/outside queries...
Single filter net ns/op: 49.8812
Batched, prepared net ns/op: 97.1514
Batched, unprepared net ns/op: 222.025
Skewed 50% in 1% net ns/op: 197.48
Skewed 80% in 20% net ns/op: 212.457
Random filter net ns/op: 226.464
Bloom comparison, for reference:
$ ./filter_bench -impl=2 -m_keys_total_max=200 -average_keys_per_filter=100000 -m_queries=50
WARNING: Assertions are enabled; benchmarks unnecessarily slow
Building...
Build avg ns/key: 35.3042
Number of filters: 1993
Total size (MB): 238.488
Reported total allocated memory (MB): 262.875
Reported internal fragmentation: 10.2255%
Bits/key stored: 10.0029
Prelim FP rate %: 0.965327
----------------------------
Mixed inside/outside queries...
Single filter net ns/op: 9.09931
Batched, prepared net ns/op: 34.21
Batched, unprepared net ns/op: 88.8564
Skewed 50% in 1% net ns/op: 139.75
Skewed 80% in 20% net ns/op: 181.264
Random filter net ns/op: 173.88
Reviewed By: jay-zhuang
Differential Revision: D26378710
Pulled By: pdillinger
fbshipit-source-id: 058428967c55ed763698284cd3b4bbe3351b6e69
Summary:
This PR adds the foundation classes for key-value integrity protection and the first use case: protecting live updates from the source buffers added to `WriteBatch` through the destination buffer in `MemTable`. The width of the protection info is not yet configurable -- only eight bytes per key is supported. This PR allows users to enable protection by constructing `WriteBatch` with `protection_bytes_per_key == 8`. It does not yet expose a way for users to get integrity protection via other write APIs (e.g., `Put()`, `Merge()`, `Delete()`, etc.).
The foundation classes (`ProtectionInfo.*`) embed the coverage info in their type, and provide `Protect.*()` and `Strip.*()` functions to navigate between types with different coverage. For making bytes per key configurable (for powers of two up to eight) in the future, these classes are templated on the unsigned integer type used to store the protection info. That integer contains the XOR'd result of hashes with independent seeds for all covered fields. For integer fields, the hash is computed on the raw unadjusted bytes, so the result is endian-dependent. The most significant bytes are truncated when the hash value (8 bytes) is wider than the protection integer.
When `WriteBatch` is constructed with `protection_bytes_per_key == 8`, we hold a `ProtectionInfoKVOTC` (i.e., one that covers key, value, optype aka `ValueType`, timestamp, and CF ID) for each entry added to the batch. The protection info is generated from the original buffers passed by the user, as well as the original metadata generated internally. When writing to memtable, each entry is transformed to a `ProtectionInfoKVOTS` (i.e., dropping coverage of CF ID and adding coverage of sequence number), since at that point we know the sequence number, and have already selected a memtable corresponding to a particular CF. This protection info is verified once the entry is encoded in the `MemTable` buffer.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7748
Test Plan:
- an integration test to verify a wide variety of single-byte changes to the encoded `MemTable` buffer are caught
- add to stress/crash test to verify it works in variety of configs/operations without intentional corruption
- [deferred] unit tests for `ProtectionInfo.*` classes for edge cases like KV swap, `SliceParts` and `Slice` APIs are interchangeable, etc.
Reviewed By: pdillinger
Differential Revision: D25754492
Pulled By: ajkr
fbshipit-source-id: e481bac6c03c2ab268be41359730f1ceb9964866
Summary:
Removed the uses of the Legacy FileWrapper classes from the source code. The wrappers were creating an additional layer of indirection/wrapping, as the Env already has a FileSystem.
Moved the Custom FileWrapper classes into the CustomEnv, as these classes are really for the private use the the CustomEnv class.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7851
Reviewed By: anand1976
Differential Revision: D26114816
Pulled By: mrambacher
fbshipit-source-id: db32840e58d969d3a0fa6c25aaf13d6dcdc74150
Summary:
Closes https://github.com/facebook/rocksdb/issues/7035
Changed how build_version.cc was generated:
- Included the GIT tag/branch in the build_version file
- Changed the "Build Date" to be:
- If the GIT branch is "clean" (no changes), the date of the last git commit
- If the branch is not clean, the current date
- Added APIs to access the "build information", rather than accessing the strings directly.
The build_version.cc file is now regenerated whenever the library objects are rebuilt.
Verified that the built files remain the same size across builds on a "clean build" and the same information is reported by sst_dump --version
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7866
Reviewed By: pdillinger
Differential Revision: D26086565
Pulled By: mrambacher
fbshipit-source-id: 6fcbe47f6033989d5cf26a0ccb6dfdd9dd239d7f
Summary:
Introduces and uses a SystemClock class to RocksDB. This class contains the time-related functions of an Env and these functions can be redirected from the Env to the SystemClock.
Many of the places that used an Env (Timer, PerfStepTimer, RepeatableThread, RateLimiter, WriteController) for time-related functions have been changed to use SystemClock instead. There are likely more places that can be changed, but this is a start to show what can/should be done. Over time it would be nice to migrate most (if not all) of the uses of the time functions from the Env to the SystemClock.
There are several Env classes that implement these functions. Most of these have not been converted yet to SystemClock implementations; that will come in a subsequent PR. It would be good to unify many of the Mock Timer implementations, so that they behave similarly and be tested similarly (some override Sleep, some use a MockSleep, etc).
Additionally, this change will allow new methods to be introduced to the SystemClock (like https://github.com/facebook/rocksdb/issues/7101 WaitFor) in a consistent manner across a smaller number of classes.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7858
Reviewed By: pdillinger
Differential Revision: D26006406
Pulled By: mrambacher
fbshipit-source-id: ed10a8abbdab7ff2e23d69d85bd25b3e7e899e90
Summary:
This provides a workaround for two race conditions that will be fixed in
a more sophisticated way later. This PR:
(1) Makes the client serialize calls to `Timer::Start()` and `Timer::Shutdown()` (see https://github.com/facebook/rocksdb/issues/7711). The long-term fix will be to make those functions thread-safe.
(2) Makes `PeriodicWorkScheduler` atomically add/cancel work together with starting/shutting down its `Timer`. The long-term fix will be for `Timer` API to offer more specialized APIs so the client will not need to synchronize.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7888
Test Plan: ran the repro provided in https://github.com/facebook/rocksdb/issues/7881
Reviewed By: jay-zhuang
Differential Revision: D25990891
Pulled By: ajkr
fbshipit-source-id: a97fdaebbda6d7db7ddb1b146738b68c16c5be38
Summary:
- Completed the switch statement for all possible `Code` values (the only one missing was `kCompactionTooLarge`).
- Removed the default case so compiler can alert us if a new value is added to `Code` without handling it in `Status::ToString()`.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7872
Test Plan:
verified the log message for this scenario looks right
```
2021/01/15-17:26:34.564450 7fa6845fe700 [ERROR] [/db_impl/db_impl_compaction_flush.cc:2621] Waiting after background compaction error: Compaction too large: , Accumulated background error counts: 1
```
Reviewed By: ramvadiv
Differential Revision: D25934539
Pulled By: ajkr
fbshipit-source-id: 2e0b3c0d993e356a4987276d6f8a163f0ee8be7a
Summary:
Change the StringEnv and related classes to be based on FileSystem APIs rather than the corresponding Env ones. The StringSink and StringSource classes were changed to be based on the corresponding FS file classes.
Part of a cleanup to use the newer interfaces. This change also eliminates some of the casts/wrappers to LegacyFile classes.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7786
Reviewed By: jay-zhuang
Differential Revision: D25761460
Pulled By: anand1976
fbshipit-source-id: 428ae8e32b3db97dbeeca08c9d3bb0d9d4d3a38f
Summary:
The test was flaky because the BG threads could increase
`running_count_` up to `job_count_` before applying their thread status
updates. Then the test thread would see non-deterministic results when
counting threads with each status. The fix is to acquire mutex in test
thread so it sees `running_count_` and thread status updated atomically.
I think simply reordering the two updates would have been insufficient
since the thread status update uses `memory_order_relaxed`. This change
happens to also eliminate an undesirable sleep loop.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7825
Test Plan:
injected sleeps to verify the failure repros before this PR and does not
repro after.
Reviewed By: jay-zhuang
Differential Revision: D25742409
Pulled By: ajkr
fbshipit-source-id: 926a2223fe856e20bc4c0c27df6736ee5cb02c97
Summary:
Added "no-elide-constructors to the ASSERT_STATUS_CHECK builds. This flag gives more errors/warnings for some of the Status checks where an inner class checks a Status and later returns it. In this case, without the elide check on, the returned status may not have been checked in the caller, thereby bypassing the checked code.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7798
Reviewed By: jay-zhuang
Differential Revision: D25680451
Pulled By: pdillinger
fbshipit-source-id: c3f14ed9e2a13f0a8c54d839d5fb4d1fc1e93917
Summary:
Primarily this change refactors the optimize_filters_for_memory
code for Bloom filters, based on malloc_usable_size, to also work for
Ribbon filters.
This change also replaces the somewhat slow but general
BuiltinFilterBitsBuilder::ApproximateNumEntries with
implementation-specific versions for Ribbon (new) and Legacy Bloom
(based on a recently deleted version). The reason is to emphasize
speed in ApproximateNumEntries rather than 100% accuracy.
Justification: ApproximateNumEntries (formerly CalculateNumEntry) is
only used by RocksDB for range-partitioned filters, called each time we
start to construct one. (In theory, it should be possible to reuse the
estimate, but the abstractions provided by FilterPolicy don't really
make that workable.) But this is only used as a heuristic estimate for
hitting a desired partitioned filter size because of alignment to data
blocks, which have various numbers of unique keys or prefixes. The two
factors lead us to prioritize reasonable speed over 100% accuracy.
optimize_filters_for_memory adds extra complication, because precisely
calculating num_entries for some allowed number of bytes depends on state
with optimize_filters_for_memory enabled. And the allocator-agnostic
implementation of optimize_filters_for_memory, using malloc_usable_size,
means we would have to actually allocate memory, many times, just to
precisely determine how many entries (keys) could be added and stay below
some size budget, for the current state. (In a draft, I got this
working, and then realized the balance of speed vs. accuracy was all
wrong.)
So related to that, I have made CalculateSpace, an internal-only API
only used for testing, non-authoritative also if
optimize_filters_for_memory is enabled. This simplifies some code.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7774
Test Plan:
unit test updated, and for FilterSize test, range of tested
values is greatly expanded (still super fast)
Also tested `db_bench -benchmarks=fillrandom,stats -bloom_bits=10 -num=1000000 -partition_index_and_filters -format_version=5 [-optimize_filters_for_memory] [-use_ribbon_filter]` with temporary debug output of generated filter sizes.
Bloom+optimize_filters_for_memory:
1 Filter size: 197 (224 in memory)
134 Filter size: 3525 (3584 in memory)
107 Filter size: 4037 (4096 in memory)
Total on disk: 904,506
Total in memory: 918,752
Ribbon+optimize_filters_for_memory:
1 Filter size: 3061 (3072 in memory)
110 Filter size: 3573 (3584 in memory)
58 Filter size: 4085 (4096 in memory)
Total on disk: 633,021 (-30.0%)
Total in memory: 634,880 (-30.9%)
Bloom (no offm):
1 Filter size: 261 (320 in memory)
1 Filter size: 3333 (3584 in memory)
240 Filter size: 3717 (4096 in memory)
Total on disk: 895,674 (-1% on disk vs. +offm; known tolerable overhead of offm)
Total in memory: 986,944 (+7.4% vs. +offm)
Ribbon (no offm):
1 Filter size: 2949 (3072 in memory)
1 Filter size: 3381 (3584 in memory)
167 Filter size: 3701 (4096 in memory)
Total on disk: 624,397 (-30.3% vs. Bloom)
Total in memory: 690,688 (-30.0% vs. Bloom)
Note that optimize_filters_for_memory is even more effective for Ribbon filter than for cache-local Bloom, because it can close the unused memory gap even tighter than Bloom filter, because of 16 byte increments for Ribbon vs. 64 byte increments for Bloom.
Reviewed By: jay-zhuang
Differential Revision: D25592970
Pulled By: pdillinger
fbshipit-source-id: 606fdaa025bb790d7e9c21601e8ea86e10541912
Summary:
Deprecate CalculateNumEntry and replace with
ApproximateNumEntries (better name) using size_t instead of int and
uint32_t, to minimize confusing casts and bad overflow behavior
(possible though probably not realistic). Bloom sizes are now explicitly
capped at max size supported by implementations: just under 4GiB for
fv=5 Bloom, and just under 512MiB for fv<5 Legacy Bloom. This
hardening could help to set up for fuzzing.
Also, since RocksDB only uses this information as an approximation
for trying to hit certain sizes for partitioned filters, it's more important
that the function be reasonably fast than for it to be completely
accurate. It's hard enough to be 100% accurate for Ribbon (currently
reversing CalculateSpace) that adding optimize_filters_for_memory
into the mix is just not worth trying to be 100% accurate for num
entries for bytes.
Also:
- Cleaned up filter_policy.h to remove MSVC warning handling and
potentially unsafe use of exception for "not implemented"
- Correct the number of entries limit beyond which current Ribbon
implementation falls back on Bloom instead.
- Consistently use "num_entries" rather than "num_entry"
- Remove LegacyBloomBitsBuilder::CalculateNumEntry as it's essentially
obsolete from general implementation
BuiltinFilterBitsBuilder::CalculateNumEntries.
- Fix filter_bench to skip some tests that don't make sense when only
one or a small number of filters has been generated.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7726
Test Plan:
expanded existing unit tests for CalculateSpace /
ApproximateNumEntries. Also manually used filter_bench to verify Legacy and
fv=5 Bloom size caps work (much too expensive for unit test). Note that
the actual bits per key is below requested due to space cap.
$ ./filter_bench -impl=0 -bits_per_key=20 -average_keys_per_filter=256000000 -vary_key_count_ratio=0 -m_keys_total_max=256 -allow_bad_fp_rate
...
Total size (MB): 511.992
Bits/key stored: 16.777
...
$ ./filter_bench -impl=2 -bits_per_key=20 -average_keys_per_filter=2000000000 -vary_key_count_ratio=0 -m_keys_total_max=2000
...
Total size (MB): 4096
Bits/key stored: 17.1799
...
$
Reviewed By: jay-zhuang
Differential Revision: D25239800
Pulled By: pdillinger
fbshipit-source-id: f94e6d065efd31e05ec630ae1a82e6400d8390c4
Summary:
To build on FreeBSD, arch_ppc_probe needs to be adapted to FreeBSD.
Since FreeBSD uses elf_aux_info as an getauxval equivalent, use it and include necessary headers:
- machine/cpu.h for PPC_FEATURE2_HAS_VEC_CRYPTO,
- sys/auxv.h for elf_aux_info,
- sys/elf_common.h for AT_HWCAP2.
elf_aux_info isn't checked for being available, because it's available since FreeBSD 12.0. rocksdb assumes using Clang on FreeBSD, but powerpc* platforms switch to Clang only since 13.0.
This patch makes rocksdb build on FreeBSD on powerpc64 and powerpc64le platforms.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7732
Reviewed By: ltamasi
Differential Revision: D25399194
Pulled By: pdillinger
fbshipit-source-id: 9c905147d75f98cd2557dd2f86a940b8e6c5afcd
Summary:
Closes - https://github.com/facebook/rocksdb/issues/7710
I tested this on an Apple DTK (Developer Transition Kit) with an Apple A12Z Bionic CPU and macOS Big Sur (11.0.1).
Previously the arm64 specific CRC optimisations were limited to Linux only OS... Well now Apple Silicon is also arm64 but runs macOS ;-)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7714
Reviewed By: ltamasi
Differential Revision: D25287349
Pulled By: pdillinger
fbshipit-source-id: 639b168bf0ac2652907531e9604936ac4974b577
Summary:
The minimum rate check in RateLimiterTest.Rate can fail in
Facebook's CI system Sandcastle, presumably due to heavily loaded
machines. This change disables the minimum rate check for Sandcastle
runs, and cleans up the code disabling it on other CI environments. (The
amount of conditionally compiled code shall be minimized.)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7728
Test Plan: try new test with and without setting envvar SANDCASTLE=1
Reviewed By: ltamasi
Differential Revision: D25247642
Pulled By: pdillinger
fbshipit-source-id: d786233af37af9a874adbb3a9e2707ec52c27a5a
Summary:
Closes https://github.com/facebook/rocksdb/issues/7691
The optimised CRC code for PPC64le which was originally imported in https://github.com/facebook/rocksdb/pull/2353 is not compatible with Clang 11. It looks like the code most likely originated from https://github.com/antonblanchard/crc32-vpmsum.
The code relied on a GCC header file `ppc-asm.h` which is not available in Clang.
To solve this, I have taken the same approach as the the upstream project from which the CRC code came ffc8018efc (diff-ec3e62c56fbcddeb07230f2a4673c1abd7f0f1cc8e48a2aa560056cfc1b25d60) and simply imported a copy of the GCC header file into our code-base which will be used when Clang is the compiler on pcc64le.
**NOTE**: The new file `util/ppc-asm.h` may have licensing implications which I guess need to be approved by RocksDB/Facebook before this is merged
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7713
Reviewed By: jay-zhuang
Differential Revision: D25222645
Pulled By: pdillinger
fbshipit-source-id: e3fec9136f26ce1eb7a027048bcf77a6cb3c769c
