Summary:
Currently it's easy to use a ton of memory with many small OptimisticTransactionDB instances, because each one by default allocates a million mutexes (40 bytes each on my compiler) for validating transactions. It even puts a lot of pressure on the allocator by allocating each one individually!
In this change:
* Create a new object and option that enables sharing these buckets of mutexes between instances. This is generally good for load balancing potential contention as various DBs become hotter or colder with txn writes. About the only cases where this sharing wouldn't make sense (e.g. each DB usually written by one thread) are cases that would be better off with OccValidationPolicy::kValidateSerial which doesn't use the buckets anyway.
* Allocate the mutexes in a contiguous array, for efficiency
* Add an option to ensure the mutexes are cache-aligned. In several other places we use cache-aligned mutexes but OptimisticTransactionDB historically does not. It should be a space-time trade-off the user can choose.
* Provide some visibility into the memory used by the mutex buckets with an ApproximateMemoryUsage() function (also used in unit testing)
* Share code with other users of "striped" mutexes, appropriate refactoring for customization & efficiency (e.g. using FastRange instead of modulus)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11439
Test Plan: unit tests added. Ran sized-up versions of stress test in unit test, including a before-and-after performance test showing no consistent difference. (NOTE: OptimisticTransactionDB not currently covered by db_stress!)
Reviewed By: ltamasi
Differential Revision: D45796393
Pulled By: pdillinger
fbshipit-source-id: ae2b3a26ad91ceeec15debcdc63ff48df6736a54
Summary:
Add the following missing options to `src/main/java/org/rocksdb/CompactRangeOptions.java` and in `java/rocksjni/options.cc` in RocksJava.
For the descriptions and API see the C++ file `include/rocksdb/options.h`, specifically the struct `CompactRangeOptions`
* full_history_ts_low
* canceled
We changed the handle to return an object (of class `Java_org_rocksdb_CompactRangeOptions`) containing a `ROCKSDB_NAMESPACE::CompactRangeOptions` at (almost certainly) 0-offset, rather than a raw `ROCKSDB_NAMESPACE::CompactRangeOptions`.
The `Java_org_rocksdb_CompactRangeOptions` contains as supplementary fields objects (std::string, std::atomic<bool>) which are passed as pointers to the `ROCKSDB_NAMESPACE::CompactRangeOptions` and which must therefore live for as long as the `ROCKSDB_NAMESPACE::CompactRangeOptions`. By placing them in a `Java_org_rocksdb_CompactRangeOptions` we achieve this.
Because the field offset of the `ROCKSDB_NAMESPACE::CompactRangeOptions` member is (very probably) 0, casting the handle to ROCKSDB_NAMESPACE::CompactRangeOptions works (i.e. old methods didn’t have to be changed), but really that’s a minefield and the correct answer is to cast to the correct type (Java_org_rocksdb_CompactRangeOptions) and then use the ROCKSDB_NAMESPACE::CompactRangeOptions field in that. So the get/set methods for existing parameters have this change.
Testing
-------
We added unit tests for getting and setting the newly implemented fields to `CompactRangeOptionsTest`
Pull Request resolved: https://github.com/facebook/rocksdb/pull/10880
Reviewed By: ajkr
Differential Revision: D41482476
Pulled By: anand1976
fbshipit-source-id: c70795e790436fb3544655920adf6fca62ed34e2
Summary:
Do not bother comparing the version of the local super version handle with the global one.
An inequality comparison result indicates nothing but a spurious obsoleteness. It only happens when the writer has increased the `ColumnFamilyData::super_version_number_`(5fc57eec2b/db/column_family.cc (L1309)) but has not yet called `ResetThreadLocalSuperVersions()`(5fc57eec2b/db/column_family.cc (L1328)) at the time when a reader reads the local handle(`void* ptr = local_sv_->Swap(SuperVersion::kSVInUse);`). In other words, the existence of a local handle is a sufficent evidence of its fressness.
With this PR, we save one or even two atomic instructions when getting a handle of super version.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11452
Reviewed By: ajkr
Differential Revision: D46059317
Pulled By: cbi42
fbshipit-source-id: 68b4b1ca8a9929a4aa470105c37a09e0625b014d
Summary:
- add TEST_TMPDIR_EXPECTED env in crash test if expected dir is on different filesystem. If TEST_TMPDIR_EXPECTED is not specified, it'll fallback to default value of TEST_TMPDIR (Same as before)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11448
Test Plan: Ran locally
Reviewed By: anand1976
Differential Revision: D45870268
Pulled By: akankshamahajan15
fbshipit-source-id: 52a7b961d3647dde023dcf7f20341558e8a5b528
Summary:
Add a util method `HandleWriteBatchTimestampSizeDifference` to handle a `WriteBatch` read from WAL log when user-defined timestamp size record is written and read. Two check modes are added: `kVerifyConsistency` that just verifies the recorded timestamp size are consistent with the running ones. This mode is to be used by `db_impl_secondary` for opening a DB as secondary instance. It will also be used by `db_impl_open` before the user comparator switch support is added to make a column switch between enabling/disable UDT feature. The other mode `kReconcileInconsistency` will be used by `db_impl_open` later when user comparator can be changed.
Another change is to extract a method `CollectColumnFamilyIdsFromWriteBatch` in db_secondary_impl.h into its standalone util file so it can be shared.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11451
Test Plan:
```
make check
./udt_util_test
```
Reviewed By: ltamasi
Differential Revision: D45894386
Pulled By: jowlyzhang
fbshipit-source-id: b96790777f154cddab6d45d9ba2e5d20ebc6fe9d
Summary:
This patch remove the "stress" aspect from the WriteUnpreparedStressTest and leave it to be a unit test for some correctness testing w.r.t. snapshot functionality. I added some read-your-write verification to the transaction test in db_stress.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11424
Test Plan:
`./write_unprepared_transaction_test`
`./db_crashtest.py whitebox --txn`
`./db_crashtest.py blackbox --txn`
Reviewed By: hx235
Differential Revision: D45551521
Pulled By: jowlyzhang
fbshipit-source-id: 20c3d510eb4255b08ddd7b6c85bdb4945436f6e8
Summary:
jemalloc was not building on M1 Macs. This makes it work.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11257
Reviewed By: anand1976
Differential Revision: D45959570
Pulled By: ajkr
fbshipit-source-id: 08c2b81b399f5003a2c159d037f9bcc5d0059556
Summary:
We want to know more about opportunities for better range filters, and the effectiveness of our own range filters. Currently the stats are very limited, essentially logging just hits and misses against prefix filters for range scans in BLOOM_FILTER_PREFIX_* without tracking the false positive rate. Perhaps confusingly, when prefix filters are used for point queries, the stats are currently going into the non-PREFIX tickers.
This change does several things:
* Introduce new stat tickers for seeks and related filtering, \*LEVEL_SEEK\*
* Most importantly, allows us to see opportunities for range filtering. Specifically, we can count how many times a seek in an SST file accesses at least one data block, and how many times at least one value() is then accessed. If a data block was accessed but no value(), we can generally assume that the key(s) seen was(were) not of interest so could have been filtered with the right kind of filter, avoiding the data block access.
* We can get the same level of detail when a filter (for now, prefix Bloom/ribbon) is used, or not. Specifically, we can infer a false positive rate for prefix filters (not available before) from the seek "false positive" rate: when a data block is accessed but no value() is called. (There can be other explanations for a seek false positive, but in typical iterator usage it would indicate a filter false positive.)
* For efficiency, I wanted to avoid making additional calls to the prefix extractor (or key comparisons, etc.), which would be required if we wanted to more precisely detect filter false positives. I believe that instrumenting value() is the best balance of efficiency vs. accurately measuring what we are often interested in.
* The stats are divided between last level and non-last levels, to help understand potential tiered storage use cases.
* The old BLOOM_FILTER_PREFIX_* stats have a different meaning: no longer referring to iterators but to point queries using prefix filters. BLOOM_FILTER_PREFIX_TRUE_POSITIVE is added for computing the prefix false positive rate on point queries, which can be due to filter false positives as well as different keys with the same prefix.
* Similarly, the non-PREFIX BLOOM_FILTER stats are now for whole key filtering only.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11460
Test Plan:
unit tests updated, including updating many to pop the stat value since last read to improve test
readability and maintainability.
Performance test shows a consistent small improvement with these changes, both with clang and with gcc. CPU profile indicates that RecordTick is using less CPU, and this makes sense at least for a high filter miss rate. Before, we were recording two ticks per filter miss in iterators (CHECKED & USEFUL) and now recording just one (FILTERED).
Create DB with
```
TEST_TMPDIR=/dev/shm ./db_bench -benchmarks=fillrandom -num=10000000 -disable_wal=1 -write_buffer_size=30000000 -bloom_bits=8 -compaction_style=2 -fifo_compaction_max_table_files_size_mb=10000 -fifo_compaction_allow_compaction=0 -prefix_size=8
```
And run simultaneous before&after with
```
TEST_TMPDIR=/dev/shm ./db_bench -readonly -benchmarks=seekrandom[-X1000] -num=10000000 -bloom_bits=8 -compaction_style=2 -fifo_compaction_max_table_files_size_mb=10000 -fifo_compaction_allow_compaction=0 -prefix_size=8 -seek_nexts=1 -duration=20 -seed=43 -threads=8 -cache_size=1000000000 -statistics
```
Before: seekrandom [AVG 275 runs] : 189680 (± 222) ops/sec; 18.4 (± 0.0) MB/sec
After: seekrandom [AVG 275 runs] : 197110 (± 208) ops/sec; 19.1 (± 0.0) MB/sec
Reviewed By: ajkr
Differential Revision: D46029177
Pulled By: pdillinger
fbshipit-source-id: cdace79a2ea548d46c5900b068c5b7c3a02e5822
Summary:
Add two type aliases for Cache: BlockCache and GeneralCache, and add LRUCacheOptions::MakeSharedGeneralCache(). This will ease upgrade to an intended future change to separate the cache API between block cache and other (general) uses, including row cache. Separating the APIs will make it easier to expose more details of block caching for customization. For example, it would be nice to pass the file unique ID and offset as the logical cache key instead of using a Slice, which could facilitate some file-specific customizations in block cache. This would also make it clear that HyperClockCache is not usable as a general cache, because it can only deal with fixed-size block cache keys.
block_cache, row_cache, and blob_cache are the uses of Cache in the public API. blob_cache should be able to use BlockCache while row_cache is a GeneralCache user, as its keys are of arbitrary size.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11450
Test Plan: see updated unit test.
Reviewed By: anand1976
Differential Revision: D45882067
Pulled By: pdillinger
fbshipit-source-id: ff5d9f0b644f87ae337a29a7728ce3ed07b2a4b2
Summary:
This PR is part of the request https://github.com/facebook/rocksdb/issues/11317.
(Another part is https://github.com/facebook/rocksdb/pull/11378)
ClipDB() will clip the entries in the CF according to the range [begin_key, end_key). All the entries outside this range will be completely deleted (including tombstones).
This feature is mainly used to ensure that there is no overlapping Key when calling CreateColumnFamilyWithImports() to import multiple CFs.
When Calling ClipDB [begin, end), there are the following steps
1. Quickly and directly delete files without overlap
DeleteFilesInRanges(nullptr, begin) + DeleteFilesInRanges(end, nullptr)
2. Delete the Key outside the range
Delete[smallest_key, begin) + Delete[end, largest_key]
3. Delete the tombstone through Manul Compact
CompactRange(option, nullptr, nullptr)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11379
Reviewed By: ajkr
Differential Revision: D45840358
Pulled By: cbi42
fbshipit-source-id: 54152e8a45fd8ede137f99787eb252f0b51440a4
Summary:
**Context/Summary:**
https://github.com/facebook/rocksdb/pull/11424 made me realize there are a couple gaps in my `ExpectedValue` comments so I updated them, along with separating `ExpectedValue` into separate files so it's clearer that `ExpectedValue` can be used without updating `ExpectedState` (e.g, TestMultiGet() where we care about value base of expected value but not updating the ExpectedState).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11456
Test Plan: CI
Reviewed By: jowlyzhang
Differential Revision: D45965070
Pulled By: hx235
fbshipit-source-id: dcee690c13b00a3119757ea9d43b646f9644e1a9
Summary:
Apply a small (and automatic) set of IntelliJ Java inspections/repairs to the Java interface to RocksDB Java and its tests.
Partly enabled by the fact that we now (from RocksDB7) require java 8.
Explicit <p> in empty lines in javadoc comments.
Parameters and variables made final where possible.
Anonymous subclasses converted lambdas.
Some tests which previously used other assertion models were converted to assertj, e.g. (assertThat(actual).isEqualTo(expected)
In a very few cases tests were found to be inoperative or broken, and were repaired. No problems with actual RocksDB behaviour were observed.
This PR is intended to replace https://github.com/facebook/rocksdb/pull/9618 - that PR was not merged, and attempts to rebase it have yielded a questionable looking diff, so we choose to go back to square 1 here, and implement a conservative set of changes.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/10951
Reviewed By: anand1976
Differential Revision: D45057849
Pulled By: ajkr
fbshipit-source-id: e4ea46bfc80518ae86f37702b03ca9352bc11c3d
Summary:
Context:
In pull request https://github.com/facebook/rocksdb/issues/11436, we are introducing a new public API `waitForCompact(const WaitForCompactOptions& wait_for_compact_options)`. This API invokes the internal implementation `waitForCompact(bool wait_unscheduled=false)`. The unscheduled parameter indicates the compactions that are not yet scheduled but are required to process items in the queue.
In certain cases, we are unable to wait for compactions, such as during a shutdown or when background jobs are paused. It is important to return the appropriate status in these scenarios. For all other cases, we should wait for all compaction and flush jobs, including the unscheduled ones. The primary purpose of this new API is to wait until the system has resolved its compaction debt. Currently, the usage of `wait_unscheduled` is limited to test code.
This pull request eliminates the usage of wait_unscheduled. The internal `waitForCompact()` API now waits for unscheduled compactions unless the db is undergoing a shutdown. In the event of a shutdown, the API returns `Status::ShutdownInProgress()`.
Additionally, a new parameter, `abort_on_pause`, has been introduced with a default value of `false`. This parameter addresses the possibility of waiting indefinitely for unscheduled jobs if `PauseBackgroundWork()` was called before `waitForCompact()` is invoked. By setting `abort_on_pause` to `true`, the API will immediately return `Status::Aborted`.
Furthermore, all tests that previously called `waitForCompact(true)` have been fixed.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11443
Test Plan:
Existing tests that involve a shutdown in progress:
- DBCompactionTest::CompactRangeShutdownWhileDelayed
- DBTestWithParam::PreShutdownMultipleCompaction
- DBTestWithParam::PreShutdownCompactionMiddle
Reviewed By: pdillinger
Differential Revision: D45923426
Pulled By: jaykorean
fbshipit-source-id: 7dc93fe6a6841a7d9d2d72866fa647090dba8eae
Summary:
In IDE navigation I find it annoying that there are two statistics.h files (etc.) and often land on the wrong one. Here I migrate several headers to use the blah.h <- blah_impl.h <- blah.cc idiom. Although clang-format wants "blah.h" to be the top include for "blah.cc", I think overall this is an improvement.
No public API changes.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11408
Test Plan: existing tests
Reviewed By: ltamasi
Differential Revision: D45456696
Pulled By: pdillinger
fbshipit-source-id: 809d931253f3272c908cf5facf7e1d32fc507373
Summary:
**Context:**
Current `NonBatchedOpsStressTest` does not allow multi-thread read (i.e, Get, Iterator) and write (i.e, Put, Merge) or delete to the same key. Every read or write/delete operation will acquire lock (`GetLocksForKeyRange`) on the target key to gain exclusive access to it. This does not align with RocksDB's nature of allowing multi-thread read and write/delete to the same key, that is concurrent threads can issue read/write/delete to RocksDB without external locking. Therefore this is a gap in our testing coverage.
To close the gap, biggest challenge remains in verifying db value against expected state in presence of parallel read and write/delete. The challenge is due to read/write/delete to the db and read/write to expected state is not within one atomic operation. Therefore we may not know the exact expected state of a certain db read, as by the time we read the expected state for that db read, another write to expected state for another db write to the same key might have changed the expected state.
**Summary:**
Credited to ajkr's idea, we now solve this challenge by breaking the 32-bits expected value of a key into different parts that can be read and write to in parallel.
Basically we divide the 32-bits expected value into `value_base` (corresponding to the previous whole 32 bits but now with some shrinking in the value base range we allow), `pending_write` (i.e, whether there is an ongoing concurrent write), `del_counter` (i.e, number of times a value has been deleted, analogous to value_base for write), `pending_delete` (similar to pending_write) and `deleted` (i.e whether a key is deleted).
Also, we need to use incremental `value_base` instead of random value base as before because we want to control the range of value base a correct db read result can possibly be in presence of parallel read and write. In that way, we can verify the correctness of the read against expected state more easily. This is at the cost of reducing the randomness of the value generated in NonBatchedOpsStressTest we are willing to accept.
(For detailed algorithm of how to use these parts to infer expected state of a key, see the PR)
Misc: hide value_base detail from callers of ExpectedState by abstracting related logics into ExpectedValue class
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11058
Test Plan:
- Manual test of small number of keys (i.e, high chances of parallel read and write/delete to same key) with equally distributed read/write/deleted for 30 min
```
python3 tools/db_crashtest.py --simple {blackbox|whitebox} --sync_fault_injection=1 --skip_verifydb=0 --continuous_verification_interval=1000 --clear_column_family_one_in=0 --max_key=10 --column_families=1 --threads=32 --readpercent=25 --writepercent=25 --nooverwritepercent=0 --iterpercent=25 --verify_iterator_with_expected_state_one_in=1 --num_iterations=5 --delpercent=15 --delrangepercent=10 --range_deletion_width=5 --use_merge={0|1} --use_put_entity_one_in=0 --use_txn=0 --verify_before_write=0 --user_timestamp_size=0 --compact_files_one_in=1000 --compact_range_one_in=1000 --flush_one_in=1000 --get_property_one_in=1000 --ingest_external_file_one_in=100 --backup_one_in=100 --checkpoint_one_in=100 --approximate_size_one_in=0 --acquire_snapshot_one_in=100 --use_multiget=0 --prefixpercent=0 --get_live_files_one_in=1000 --manual_wal_flush_one_in=1000 --pause_background_one_in=1000 --target_file_size_base=524288 --write_buffer_size=524288 --verify_checksum_one_in=1000 --verify_db_one_in=1000
```
- Rehearsal stress test for normal parameter and aggressive parameter to see if such change can find what existing stress test can find (i.e, no regression in testing capability)
- [Ongoing]Try to find new bugs with this change that are not found by current NonBatchedOpsStressTest with no parallel read and write/delete to same key
Reviewed By: ajkr
Differential Revision: D42257258
Pulled By: hx235
fbshipit-source-id: e6fdc18f1fad3753e5ac91731483a644d9b5b6eb
Summary:
When the DB is opened, RocksDB creates a temp OPTIONS file, writes the current options to it, and renames it. In case of a failure, the temp file is left behind, and is not deleted by PurgeObsoleteFiles(). Fix this by explicitly deleting the temp file if writing to it or renaming it fails.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11423
Test Plan: Add a unit test
Reviewed By: akankshamahajan15
Differential Revision: D45540454
Pulled By: anand1976
fbshipit-source-id: 47facdc30d8cc5667036312d04b21d3fc253c92e
Summary:
Added a ticker stat, `BLOCK_CHECKSUM_MISMATCH_COUNT`, to count how many block checksum verifications detected a mismatch.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11438
Test Plan: new unit test
Reviewed By: pdillinger
Differential Revision: D45788179
Pulled By: ajkr
fbshipit-source-id: e2b44eba7c23b3e110ebe69eaa78a710dec2590f
Summary:
This patch adds support to write and read a user-defined timestamp size record in log writer and log reader. It will be used by WAL logs to persist the user-defined timestamp format for subsequent WriteBatch records. Reading and writing UDT sizes for WAL logs are not included in this patch. It will be in a follow up.
The syntax for the record is: at write time, one such record is added when log writer encountered any non-zero UDT size it hasn't recorded so far. At read time, all such records read up to a point are accumulated and applicable to all subsequent WriteBatch records.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11433
Test Plan:
```
make clean && make -j32 all
./log_test --gtest_filter="*WithTimestampSize*"
```
Reviewed By: ltamasi
Differential Revision: D45678708
Pulled By: jowlyzhang
fbshipit-source-id: b770c8f45bb7b9383b14aac9f22af781304fb41d
Summary:
- Add a new option `CompactionOptionsFIFO::file_temperature_age_thresholds` that allows user to specify age thresholds for compacting files to different temperatures. File temperature can be used to store files in different storage media. The new options allows specifying multiple temperature-age pairs. The option uses struct for a temperature-age pair to use the existing parsing functionality to make the option dynamically settable.
- Deprecate the old option `age_for_warm` that was added for a similar purpose.
- Compaction score calculation logic is updated to check if a file needs to be compacted to change its temperature.
- Some refactoring is done in `FIFOCompactionPicker::PickTemperatureChangeCompaction`.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11428
Test Plan: adapted unit tests that were for `age_for_warm` to this new option.
Reviewed By: ajkr
Differential Revision: D45611412
Pulled By: cbi42
fbshipit-source-id: 2dc384841f61cc04abb9681e31aa2de0f0b06106
Summary:
Context:
This pull request update is in response to a comment made on https://github.com/facebook/rocksdb/pull/8596#discussion_r680264932. The current implementation of RefillBytesAndGrantRequestsLocked() drains all available_bytes, but the first request after the last wave of requesting/bytes granting is done is not being handled in the same way.
This creates a scenario where if a request for a large amount of bytes is enqueued first, but there are not enough available_bytes to fulfill it, the request is put to sleep until the next refill time. Meanwhile, a later request for a smaller number of bytes comes in and is granted immediately. This behavior is not fair as the earlier request was made first.
To address this issue, we have made changes to the code to exhaust the remaining available bytes from the request and queue the remaining. With this change, requests are granted in the order they are received, ensuring that earlier requests are not unfairly delayed by later, smaller requests. The specific scenario described above will no longer occur with this change. Also consolidated `granted` and `request_bytes` as part of the change since `granted` is equivalent to `request_bytes == 0`
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11425
Test Plan: Added `AvailableByteSizeExhaustTest`
Reviewed By: hx235
Differential Revision: D45570711
Pulled By: jaykorean
fbshipit-source-id: a7117ed17bf4b8a7ae0f76124cb41902db1a2592
Summary:
**Background** - runtime detection of certain x86 CPU features was added for optimizing CRC32c checksums, where performance is dramatically affected by the availability of certain CPU instructions and code using intrinsics for those instructions. And Java builds with native library try to be broadly compatible but performant.
What has changed is that CRC32c is no longer the most efficient cheecksum on contemporary x86_64 hardware, nor the default checksum. XXH3 is generally faster and not as dramatically impacted by the availability of certain CPU instructions. For example, on my Skylake system using db_bench (similar on an older Skylake system without AVX512):
PORTABLE=1 empty USE_SSE : xxh3->8 GB/s crc32c->0.8 GB/s (no SSE4.2 nor AVX2 instructions)
PORTABLE=1 USE_SSE=1 : xxh3->19 GB/s crc32c->16 GB/s (with SSE4.2 and AVX2)
PORTABLE=0 USE_SSE ignored: xxh3->28 GB/s crc32c->16 GB/s (also some AVX512)
Testing a ~10 year old system, with SSE4.2 but without AVX2, crc32c is a similar speed to the new systems but xxh3 is only about half that speed, also 8GB/s like the non-AVX2 compile above. Given that xxh3 has specific optimization for AVX2, I think we can infer that that crc32c is only fastest for that ~2008-2013 period when SSE4.2 was included but not AVX2. And given that xxh3 is only about 2x slower on these systems (not like >10x slower for unoptimized crc32c), I don't think we need to invest too much in optimally adapting to these old cases.
x86 hardware that doesn't support fast CRC32c is now extremely rare, so requiring a custom build to support such hardware is fine IMHO.
**This change** does two related things:
* Remove runtime CPU detection for optimizing CRC32c on x86. Maintaining this code is non-zero work, and compiling special code that doesn't work on the configured target instruction set for code generation is always dubious. (On the one hand we have to ensure the CRC32c code uses SSE4.2 but on the other hand we have to ensure nothing else does.)
* Detect CPU features in source code, not in build scripts. Although there are some hypothetical advantages to detectiong in build scripts (compiler generality), RocksDB supports at least three build systems: make, cmake, and buck. It's not practical to support feature detection on all three, and we have suffered from missed optimization opportunities by relying on missing or incomplete detection in cmake and buck. We also depend on some components like xxhash that do source code detection anyway.
**In more detail:**
* `HAVE_SSE42`, `HAVE_AVX2`, and `HAVE_PCLMUL` replaced by standard macros `__SSE4_2__`, `__AVX2__`, and `__PCLMUL__`.
* MSVC does not provide high fidelity defines for SSE, PCLMUL, or POPCNT, but we can infer those from `__AVX__` or `__AVX2__` in a compatibility header. In rare cases of false negative or false positive feature detection, a build engineer should be able to set defines to work around the issue.
* `__POPCNT__` is another standard define, but we happen to only need it on MSVC, where it is set by that compatibility header, or can be set by the build engineer.
* `PORTABLE` can be set to a CPU type, e.g. "haswell", to compile for that CPU type.
* `USE_SSE` is deprecated, now equivalent to PORTABLE=haswell, which roughly approximates its old behavior.
Notably, this change should enable more builds to use the AVX2-optimized Bloom filter implementation.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11419
Test Plan:
existing tests, CI
Manual performance tests after the change match the before above (none expected with make build).
We also see AVX2 optimized Bloom filter code enabled when expected, by injecting a compiler error. (Performance difference is not big on my current CPU.)
Reviewed By: ajkr
Differential Revision: D45489041
Pulled By: pdillinger
fbshipit-source-id: 60ceb0dd2aa3b365c99ed08a8b2a087a9abb6a70
Summary:
See motivation and description in new ShardedCacheOptions::hash_seed option.
Updated db_bench so that its seed param is used for the cache hash seed.
Made its code more safe to ensure seed is set before use.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11391
Test Plan:
unit tests added / updated
**Performance** - no discernible difference seen running cache_bench repeatedly before & after. With lru_cache and hyper_clock_cache.
Reviewed By: hx235
Differential Revision: D45557797
Pulled By: pdillinger
fbshipit-source-id: 40bf4da6d66f9d41a8a0eb8e5cf4246a4aa07934
Summary:
Fix build error: variable 'base_level' may be uninitialized
```
db_impl_compaction_flush.cc:1195:21: error: variable 'base_level' may be uninitialized when used here [-Werror,-Wconditional-uninitialized]
level = base_level;
```
^~~~~~~~~~
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11435
Test Plan: CircleCI jobs
Reviewed By: cbi42
Differential Revision: D45708176
Pulled By: akankshamahajan15
fbshipit-source-id: 851b1205b22b63d728495e5735fa91b0ad8e012b
Summary:
**Context:**
We prefetch the tail part of a SST file (i.e, the blocks after data blocks till the end of the file) during each SST file open in hope to prefetch all the stuff at once ahead of time for later read e.g, footer, meta index, filter/index etc. The existing approach to estimate the tail size to prefetch is through `TailPrefetchStats` heuristics introduced in https://github.com/facebook/rocksdb/pull/4156, which has caused small reads in unlucky case (e.g, small read into the tail buffer during table open in thread 1 under the same BlockBasedTableFactory object can make thread 2's tail prefetching use a small size that it shouldn't) and is hard to debug. Therefore we decide to record the exact tail size and use it directly to prefetch tail of the SST instead of relying heuristics.
**Summary:**
- Obtain and record in manifest the tail size in `BlockBasedTableBuilder::Finish()`
- For backward compatibility, we fall back to TailPrefetchStats and last to simple heuristics that the tail size is a linear portion of the file size - see PR conversation for more.
- Make`tail_start_offset` part of the table properties and deduct tail size to record in manifest for external files (e.g, file ingestion, import CF) and db repair (with no access to manifest).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11406
Test Plan:
1. New UT
2. db bench
Note: db bench on /tmp/ where direct read is supported is too slow to finish and the default pinning setting in db bench is not helpful to profile # sst read of Get. Therefore I hacked the following to obtain the following comparison.
```
diff --git a/table/block_based/block_based_table_reader.cc b/table/block_based/block_based_table_reader.cc
index bd5669f0f..791484c1f 100644
--- a/table/block_based/block_based_table_reader.cc
+++ b/table/block_based/block_based_table_reader.cc
@@ -838,7 +838,7 @@ Status BlockBasedTable::PrefetchTail(
&tail_prefetch_size);
// Try file system prefetch
- if (!file->use_direct_io() && !force_direct_prefetch) {
+ if (false && !file->use_direct_io() && !force_direct_prefetch) {
if (!file->Prefetch(prefetch_off, prefetch_len, ro.rate_limiter_priority)
.IsNotSupported()) {
prefetch_buffer->reset(new FilePrefetchBuffer(
diff --git a/tools/db_bench_tool.cc b/tools/db_bench_tool.cc
index ea40f5fa0..39a0ac385 100644
--- a/tools/db_bench_tool.cc
+++ b/tools/db_bench_tool.cc
@@ -4191,6 +4191,8 @@ class Benchmark {
std::shared_ptr<TableFactory>(NewCuckooTableFactory(table_options));
} else {
BlockBasedTableOptions block_based_options;
+ block_based_options.metadata_cache_options.partition_pinning =
+ PinningTier::kAll;
block_based_options.checksum =
static_cast<ChecksumType>(FLAGS_checksum_type);
if (FLAGS_use_hash_search) {
```
Create DB
```
./db_bench --bloom_bits=3 --use_existing_db=1 --seed=1682546046158958 --partition_index_and_filters=1 --statistics=1 -db=/dev/shm/testdb/ -benchmarks=readrandom -key_size=3200 -value_size=512 -num=1000000 -write_buffer_size=6550000 -disable_auto_compactions=false -target_file_size_base=6550000 -compression_type=none
```
ReadRandom
```
./db_bench --bloom_bits=3 --use_existing_db=1 --seed=1682546046158958 --partition_index_and_filters=1 --statistics=1 -db=/dev/shm/testdb/ -benchmarks=readrandom -key_size=3200 -value_size=512 -num=1000000 -write_buffer_size=6550000 -disable_auto_compactions=false -target_file_size_base=6550000 -compression_type=none
```
(a) Existing (Use TailPrefetchStats for tail size + use seperate prefetch buffer in PartitionedFilter/IndexReader::CacheDependencies())
```
rocksdb.table.open.prefetch.tail.hit COUNT : 3395
rocksdb.sst.read.micros P50 : 5.655570 P95 : 9.931396 P99 : 14.845454 P100 : 585.000000 COUNT : 999905 SUM : 6590614
```
(b) This PR (Record tail size + use the same tail buffer in PartitionedFilter/IndexReader::CacheDependencies())
```
rocksdb.table.open.prefetch.tail.hit COUNT : 14257
rocksdb.sst.read.micros P50 : 5.173347 P95 : 9.015017 P99 : 12.912610 P100 : 228.000000 COUNT : 998547 SUM : 5976540
```
As we can see, we increase the prefetch tail hit count and decrease SST read count with this PR
3. Test backward compatibility by stepping through reading with post-PR code on a db generated pre-PR.
Reviewed By: pdillinger
Differential Revision: D45413346
Pulled By: hx235
fbshipit-source-id: 7d5e36a60a72477218f79905168d688452a4c064
Summary:
Roughly group ReadOptions into those that apply generally and those that only apply to range scans. Also use field assignment idiom to simplify specification of default values.
Also some rearranging to reduce unused padding. sizeof(ReadOptions) was 144 on my system, now 136.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11430
Test Plan: existing tests, no functional change intended
Reviewed By: hx235
Differential Revision: D45626508
Pulled By: pdillinger
fbshipit-source-id: 227d4158c5123405324f273ded2eb9d8bce86364
Summary:
This PR adds a plugin that supports AES-CTR encryption for RocksDB based on highly performant intel open-source cryptographic library IPP-Crypto.
Details:
- supports AES-128, AES-192, and AES-256.
- uses the CTR mode of operation.
- based on the Intel® crypto library -- https://github.com/intel/ipp-crypto.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11429
Reviewed By: cbi42
Differential Revision: D45622342
Pulled By: ajkr
fbshipit-source-id: 2463fa2b8ae625fdd7d83768e274c74e3f2a0f46
Summary:
Document ReadOptions::io_activity as internal-use-only. And to keep kUnknown as last (and why).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11427
Test Plan: comments only
Reviewed By: hx235
Differential Revision: D45576986
Pulled By: pdillinger
fbshipit-source-id: aae15aa22ea91370c2b7366154e45d4b91a79ad2
Summary:
the test is flaky when compiled with `make -j56 COERCE_CONTEXT_SWITCH=1 ./db_universal_compaction_test`. The cause is that a manual compaction `CompactRange()` can finish and return before obsolete files are deleted. One reason for this is that a manual compaction waits until `manual.done` is set here 62fc15f009/db/db_impl/db_impl_compaction_flush.cc (L1978)
and the compaction thread can set `manual.done`:
62fc15f009/db/db_impl/db_impl_compaction_flush.cc (L3672)
and then temporarily release mutex_:
62fc15f009/db/db_impl/db_impl_files.cc (L317)
before purging obsolete files:
62fc15f009/db/db_impl/db_impl_compaction_flush.cc (L3144)
With `COERCE_CONTEXT_SWITCH=1`, `bg_cv_.SignalAll()` is called during `mutex_.Lock()`, so the manual compaction thread can wake up and return before obsolete files are deleted. Updated the test to only count live SST files.
Also updated `FindObsoleteFiles()` to avoid locking a locked mutex.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11412
Test Plan: `make -j56 COERCE_CONTEXT_SWITCH=1 ./db_universal_compaction_test`
Reviewed By: ajkr
Differential Revision: D45342242
Pulled By: cbi42
fbshipit-source-id: 955c9796aa3f484e3557d300f97cffacb3ed9b0c
Summary:
when I use g++-13 to exec the `make all` command, the output throws the warnings.
```
db/compaction/compaction_job_test.cc: In member function ‘void rocksdb::CompactionJobTestBase::AddMockFile(const rocksdb::mock::KVVector&, int)’:
db/compaction/compaction_job_test.cc:376:57: error: redundant move in initialization [-Werror=redundant-move]
376 | env_, GenerateFileName(file_number), std::move(contents)));
| ~~~~~~~~~^~~~~~~~~~
db/compaction/compaction_job_test.cc:375:7: note: in expansion of macro ‘EXPECT_OK’
375 | EXPECT_OK(mock_table_factory_->CreateMockTable(
| ^~~~~~~~~
db/compaction/compaction_job_test.cc:376:57: note: remove ‘std::move’ call
376 | env_, GenerateFileName(file_number), std::move(contents)));
| ~~~~~~~~~^~~~~~~~~~
db/compaction/compaction_job_test.cc:375:7: note: in expansion of macro ‘EXPECT_OK’
375 | EXPECT_OK(mock_table_factory_->CreateMockTable(
| ^~~~~~~~~
cc1plus: all warnings being treated as errors
make: *** [Makefile:2507: db/compaction/compaction_job_test.o] Error 1
```
and I also add some `(void)unused_variable` statements because of the cmake argument `-Wunused-but-set-variable -Wunused-but-set-variable`
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11418
Reviewed By: akankshamahajan15
Differential Revision: D45528223
Pulled By: ajkr
fbshipit-source-id: fee1a77c30039a56b481de953f0a834cc788abbc
Summary:
When a DB is opened, RocksDB creates an empty WAL file. When the DB is reopened and the WAL is empty, the min log number to keep is not advanced until a memtable flush happens. If a process crashes soon after reopening the DB, its likely that no memtable flush would have happened, which means the empty WAL file is not deleted. In a crash loop scenario, this leads to empty WAL files accumulating. Fix this by ensuring the min log number is advanced if the WAL is empty.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11409
Test Plan: Add a unit test
Reviewed By: ajkr
Differential Revision: D45281685
Pulled By: anand1976
fbshipit-source-id: 0225877c613e65ffb30972a0051db2830105423e
Summary:
For better clarity, encouraging more options explicitly specified using fields rather than positionally via constructor parameter lists. Simplifies code maintenance as new fields are added. Deprecate some cases of the confusing pattern of NewWhatever() functions returning shared_ptr.
Net reduction of about 70 source code lines (including comments).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11386
Test Plan: existing tests
Reviewed By: ajkr
Differential Revision: D45059075
Pulled By: pdillinger
fbshipit-source-id: d53fa09b268024f9c55254bb973b6c69feebf41a
Summary:
lldb is more supported for Meta infrastructure than gdb, so adding support for it in generating stack traces and attaching debugger on crash. For now you need to set ROCKSDB_LLDB_STACK=1 for stack traces or ROCKSDB_DEBUG=lldb for interactive debugging.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11413
Test Plan: some manual testing (no production code changes)
Reviewed By: ajkr
Differential Revision: D45360952
Pulled By: pdillinger
fbshipit-source-id: 862bc8800eb03e3bdc1be8c0702960a19db45be8
Summary:
Seen in Meta-internal builds that manually depend on rocksdb_whole_archive_lib and want to automatically depend on rocksdb_lib. This change puts rocksdb_lib in the ancestry of rocksdb_whole_archive_lib, and buck2 appears to recognize that even if rocksdb_lib is listed as a separate dependency downstream.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11421
Test Plan: Run failing internal build with the change. See T147085939
Reviewed By: akankshamahajan15
Differential Revision: D45446689
Pulled By: pdillinger
fbshipit-source-id: e8a891fa020dfcf0564b35d30511d70347650fa8
Summary:
The old `StackableDB` based BlobDB implementation relies on a DB listener to track the total size of the SST files in the database and to trigger FIFO eviction. Some test cases in `BlobDBTest` assume that the listener is notified by the time `DB::Flush` returns, which is not guaranteed (side note: `TEST_WaitForFlushMemTable` would not guarantee this either). The patch fixes these tests by using `SyncPoint`s to make sure the listener is actually called before verifying the FIFO behavior.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11417
Test Plan:
```
make -j56 COERCE_CONTEXT_SWITCH=1 blob_db_test
./blob_db_test --gtest_filter=BlobDBTest.FIFOEviction_TriggerOnSSTSizeChange
./blob_db_test --gtest_filter=BlobDBTest.FilterForFIFOEviction
./blob_db_test --gtest_filter=BlobDBTest.FIFOEviction_NoEnoughBlobFilesToEvict
```
Reviewed By: ajkr
Differential Revision: D45407135
Pulled By: ltamasi
fbshipit-source-id: fcd63d76937d2c975f569a6635ce8730772a3d75
Summary:
add option `block_protection_bytes_per_key` and implementation for block per key-value checksum. The main changes are
1. checksum construction and verification in block.cc/h
2. pass the option `block_protection_bytes_per_key` around (mainly for methods defined in table_cache.h)
3. unit tests/crash test updates
Tests:
* Added unit tests
* Crash test: `python3 tools/db_crashtest.py blackbox --simple --block_protection_bytes_per_key=1 --write_buffer_size=1048576`
Follow up (maybe as a separate PR): make sure corruption status returned from BlockIters are correctly handled.
Performance:
Turning on block per KV protection has a non-trivial negative impact on read performance and costs additional memory.
For memory, each block includes additional 24 bytes for checksum-related states beside checksum itself. For CPU, I set up a DB of size ~1.2GB with 5M keys (32 bytes key and 200 bytes value) which compacts to ~5 SST files (target file size 256 MB) in L6 without compression. I tested readrandom performance with various block cache size (to mimic various cache hit rates):
```
SETUP
make OPTIMIZE_LEVEL="-O3" USE_LTO=1 DEBUG_LEVEL=0 -j32 db_bench
./db_bench -benchmarks=fillseq,compact0,waitforcompaction,compact,waitforcompaction -write_buffer_size=33554432 -level_compaction_dynamic_level_bytes=true -max_background_jobs=8 -target_file_size_base=268435456 --num=5000000 --key_size=32 --value_size=200 --compression_type=none
BENCHMARK
./db_bench --use_existing_db -benchmarks=readtocache,readrandom[-X10] --num=5000000 --key_size=32 --disable_auto_compactions --reads=1000000 --block_protection_bytes_per_key=[0|1] --cache_size=$CACHESIZE
The readrandom ops/sec looks like the following:
Block cache size: 2GB 1.2GB * 0.9 1.2GB * 0.8 1.2GB * 0.5 8MB
Main 240805 223604 198176 161653 139040
PR prot_bytes=0 238691 226693 200127 161082 141153
PR prot_bytes=1 214983 193199 178532 137013 108211
prot_bytes=1 vs -10% -15% -10.8% -15% -23%
prot_bytes=0
```
The benchmark has a lot of variance, but there was a 5% to 25% regression in this benchmark with different cache hit rates.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11287
Reviewed By: ajkr
Differential Revision: D43970708
Pulled By: cbi42
fbshipit-source-id: ef98d898b71779846fa74212b9ec9e08b7183940
Summary:
Tweak some bounds and things, and reduce risk of surprise results by running on all supported compressions (mostly).
Also improves the precise compressibility of CompressibleString by using RandomBinaryString.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11396
Test Plan: updated tests
Reviewed By: ltamasi
Differential Revision: D45211938
Pulled By: pdillinger
fbshipit-source-id: 9dc1dd8574a60a9364efe18558be66d31a35598b
Summary:
## Option API updates
* Add new CompressionOptions::max_compressed_bytes_per_kb, which corresponds to 1024.0 / min allowable compression ratio. This avoids the hard-coded minimum ratio of 8/7.
* Remove unnecessary constructor for CompressionOptions.
* Document undocumented CompressionOptions. Use idiom for default values shown clearly in one place (not precariously repeated).
## Stat API updates
* Deprecate the BYTES_COMPRESSED, BYTES_DECOMPRESSED histograms. Histograms incur substantial extra space & time costs compared to tickers, and the distribution of uncompressed data block sizes tends to be uninteresting. If we're interested in that distribution, I don't see why it should be limited to blocks stored as compressed.
* Deprecate the NUMBER_BLOCK_NOT_COMPRESSED ticker, because the name is very confusing.
* New or existing tickers relevant to compression:
* BYTES_COMPRESSED_FROM
* BYTES_COMPRESSED_TO
* BYTES_COMPRESSION_BYPASSED
* BYTES_COMPRESSION_REJECTED
* COMPACT_WRITE_BYTES + FLUSH_WRITE_BYTES (both existing)
* NUMBER_BLOCK_COMPRESSED (existing)
* NUMBER_BLOCK_COMPRESSION_BYPASSED
* NUMBER_BLOCK_COMPRESSION_REJECTED
* BYTES_DECOMPRESSED_FROM
* BYTES_DECOMPRESSED_TO
We can compute a number of things with these stats:
* "Successful" compression ratio: BYTES_COMPRESSED_FROM / BYTES_COMPRESSED_TO
* Compression ratio of data on which compression was attempted: (BYTES_COMPRESSED_FROM + BYTES_COMPRESSION_REJECTED) / (BYTES_COMPRESSED_TO + BYTES_COMPRESSION_REJECTED)
* Compression ratio of data that could be eligible for compression: (BYTES_COMPRESSED_FROM + X) / (BYTES_COMPRESSED_TO + X) where X = BYTES_COMPRESSION_REJECTED + NUMBER_BLOCK_COMPRESSION_REJECTED
* Overall SST compression ratio (compression disabled vs. actual): (Y - BYTES_COMPRESSED_TO + BYTES_COMPRESSED_FROM) / Y where Y = COMPACT_WRITE_BYTES + FLUSH_WRITE_BYTES
Keeping _REJECTED separate from _BYPASSED helps us to understand "wasted" CPU time in compression.
## BlockBasedTableBuilder
Various small refactorings, optimizations, and name clean-ups.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11388
Test Plan:
unit tests added
* `options_settable_test.cc`: use non-deprecated idiom for configuring CompressionOptions from string. The old idiom is tested elsewhere and does not need to be updated to support the new field.
Reviewed By: ajkr
Differential Revision: D45128202
Pulled By: pdillinger
fbshipit-source-id: 5a652bf5c022b7ec340cf79018cccf0686962803
Peter Dillinger