Summary:
An internal user wants to be able to dynamically switch between Bloom and Ribbon filters, without a custom FilterPolicy. Making `filter_policy` mutable would actually make issue https://github.com/facebook/rocksdb/issues/10079 worse, because it would be a race on a pointer field, not just on scalars.
As a reasonable compromise until that is fixed, I am enabling dynamic control over Bloom vs. Ribbon choice by making
RibbonFilterPolicy::bloom_before_level mutable, and doing that safely by using an atomic.
I've also slightly tweaked the interpretation of that field so that setting it to INT_MAX really means "always Bloom."
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11838
Test Plan: unit tests added/extended. crash test updated for SetOptions call and tested under TSAN with amplified probability (lower set_options_one_in).
Reviewed By: ajkr
Differential Revision: D49296284
Pulled By: pdillinger
fbshipit-source-id: e4251c077510df9a9c719876f482448c0d15402a
Summary:
- As a follow up from https://github.com/facebook/rocksdb/issues/11799, adding `Env::IOActivity::kMultiGetEntity` support to `DBImpl::MultiGetEntity()`.
## Minor Refactor
- Because both `DBImpl::MultiGet()` and `DBImpl::MultiGetEntity()` call `DBImpl::MultiGetCommon()` which later calls `DBImpl::MultiGetWithCallback()` where we check `Env::IOActivity::kMultiGet`, minor refactor was needed so that we don't check `Env::IOActivity::kMultiGet` for `DBImpl::MultiGetEntity()`.
- I still see more areas for refactoring to avoid duplicate code of checking IOActivity and setting it when Unknown, but this will be addressed separately.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11842
Test Plan:
- Added the `ThreadStatus::OperationType::OP_MULTIGETENTITY` in `db_stress` to verify the pass-down IOActivity in a thread aligns with the actual activity the thread is doing.
```
python3 tools/db_crashtest.py blackbox --max_key=25000000 --write_buffer_size=4194304 --max_bytes_for_level_base=2097152 --target_file_size_base=2097152 --periodic_compaction_seconds=0 --use_put_entity_one_in=10 --use_get_entity=1 --duration=60 --interval=10
python3 tools/db_crashtest.py blackbox --simple --max_key=25000000 --write_buffer_size=4194304 --max_bytes_for_level_base=2097152 --target_file_size_base=2097152 --periodic_compaction_seconds=0 --use_put_entity_one_in=10 --use_get_entity=1 --duration=60 --interval=10
python3 tools/db_crashtest.py blackbox --cf_consistency --max_key=25000000 --write_buffer_size=4194304 --max_bytes_for_level_base=2097152 --target_file_size_base=2097152 --periodic_compaction_seconds=0 --use_put_entity_one_in=10 --use_get_entity=1 --duration=60 --interval=10
```
Reviewed By: ltamasi
Differential Revision: D49329575
Pulled By: jaykorean
fbshipit-source-id: 05198f1d3f92e6be3d42a3d184bacb3ab2ce6923
Summary:
This PR resolves https://github.com/facebook/rocksdb/issues/10487 & https://github.com/facebook/rocksdb/issues/10536, user code needs to call Refresh() periodically.
The main code change is to support range deletions. A range tombstone iterator uses a sequence number as upper bound to decide which range tombstones are effective. During Iterator refresh, this sequence number upper bound needs to be updated for all range tombstone iterators under DBIter and LevelIterator. LevelIterator may create new table iterators and range tombstone iterator during scanning, so it needs to be aware of iterator refresh. The code path that propagates this change is `db_iter_->set_sequence(read_seq) -> MergingIterator::SetRangeDelReadSeqno() -> TruncatedRangeDelIterator::SetRangeDelReadSeqno() and LevelIterator::SetRangeDelReadSeqno()`.
This change also fixes an issue where range tombstone iterators created by LevelIterator may access ReadOptions::snapshot, even though we do not explicitly require users to keep a snapshot alive after creating an Iterator.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/10594
Test Plan:
* New unit tests.
* Add Iterator::Refresh(snapshot) to stress test. Note that this change only adds tests for refreshing to the same snapshot since this is the main target use case.
TODO in a following PR:
* Stress test Iterator::Refresh() to different snapshots or no snapshot.
Reviewed By: ajkr
Differential Revision: D48456896
Pulled By: cbi42
fbshipit-source-id: 2e642c04e91235cc9542ef4cd37b3c20823bd779
Summary:
**Context/Summary:**
It allows db bench reflect the default behavior of this option. For example, we recently changed its default value.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11831
Test Plan: No code change
Reviewed By: cbi42
Differential Revision: D49253690
Pulled By: hx235
fbshipit-source-id: 445d4e54f62b4b538626e301a3014d2f00849d30
Summary:
Application using rocksdb today dont have much control over the cost of reads when merge-ops are enabled, expect for waiting for compactions to kick in or using max_successive_merges hint, which only applies to memtable. This change adds Transaction::CollapseKey api giving applications the ability to request merge chain collapse on-demand, when they detect high read costs due to merges. Currently, this only supported on PessimisticTransactions.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11815
Test Plan: Add a unit test
Reviewed By: ajkr
Differential Revision: D49309387
Pulled By: sarangbh
fbshipit-source-id: a1eb5cc9e3bd4b3206a36150aacead770318e3e1
Summary:
Update the logic in FilePrefetchBuffer to update `upper_bound_offset_` during reseek. During Reseek, `iterate_upper_bound` can be changed dynamically. So added an API to update that in FilePrefetchBuffer.
Added unit test to confirm the behavior.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11775
Test Plan:
- Check stress tests in case there is any failure after this diff.
- make crash_test -j32 with auto_readahead_size=1 passed locally
Reviewed By: anand1976
Differential Revision: D48815177
Pulled By: akankshamahajan15
fbshipit-source-id: 5f44fbb3af06c86a1c38f139c5fa4543891837f4
Summary:
When `MultiGet` acquires `SuperVersion` via locking the db mutex and get the current `ColumnFamilyData::super_version_`, its corresponding cleanup logic is not correctly done.
It's currently doing this:
`MultiGetColumnFamilyData::cfd->GetSuperVersion().Unref()`
This operates on the most recent `SuperVersion` without locking db mutex , which is not thread safe by itself. And this unref operation is intended for the originally acquired `SuperVersion` instead of the current one. Because a race condition could happen where a new `SuperVersion` is installed in between this `MultiGet`'s ref and unref. When this race condition does happen, it's not sufficient to just unref the `SuperVersion`, `DBImpl::CleanupSuperVersion` should be called instead to properly clean up the `SuperVersion` had this `MultiGet` call be its last reference holder.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11830
Test Plan:
`make all check`
Added a unit test that would originally fail
Reviewed By: ltamasi
Differential Revision: D49287715
Pulled By: jowlyzhang
fbshipit-source-id: 8353636ee11b2e90d85c677a96a92360072644b0
Summary:
`GetEntity` API support for ReadOnly DB and Secondary DB.
- Introduced `GetImpl()` with `GetImplOptions` in `db_impl_readonly` and refactored current `Get()` logic into `GetImpl()` so that look up logic can be reused for `GetEntity()` (Following the same pattern as `DBImpl::Get()` and `DBImpl::GetEntity()`)
- Introduced `GetImpl()` with `GetImplOptions` in `db_impl_secondary` and refactored current `GetImpl()` logic. This is to make `DBImplSecondary::Get/GetEntity` consistent with `DBImpl::Get/GetEntity` and `DBImplReadOnly::Get/GetEntity`
- `GetImpl()` in `db_impl` is now virtual. both `db_impl_readonly` and `db_impl_secondary`'s `Get()` override are no longer needed since all three dbs now have the same `Get()` which calls `GetImpl()` internally.
- `GetImpl()` in `DBImplReadOnly` and `DBImplSecondary` now pass in `columns` instead of `nullptr` in lookup functions like `memtable->get()`
- Introduced `GetEntity()` API in `DBImplReadOnly` and `DBImplSecondary` which simply calls `GetImpl()` with `columns` set in `GetImplOptions`.
- Introduced `Env::IOActivity::kGetEntity` and set read_options.io_activity to `Env::IOActivity::kGetEntity` for `GetEntity()` operations (in db_impl)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11799
Test Plan:
**Unit Tests**
- Added verification in `DBWideBasicTest::PutEntity` by Reopening DB as ReadOnly with the same setup.
- Added verification in `DBSecondaryTest::ReopenAsSecondary` by calling `PutEntity()` and `GetEntity()` on top of existing `Put()` and `Get()`
- `make -j64 check`
**Crash Tests**
- `python3 tools/db_crashtest.py blackbox --max_key=25000000 --write_buffer_size=4194304 --max_bytes_for_level_base=2097152 --target_file_size_base=2097152 --periodic_compaction_seconds=0 --use_put_entity_one_in=10 --use_get_entity=1 --duration=60 --inter
val=10`
- `python3 tools/db_crashtest.py blackbox --simple --max_key=25000000 --write_buffer_size=4194304 --max_bytes_for_level_base=2097152 --target_file_size_base=2097152 --periodic_compaction_seconds=0 --use_put_entity_one_in=10 --use_get_entity=1 `
- `python3 tools/db_crashtest.py blackbox --cf_consistency --max_key=25000000 --write_buffer_size=4194304 --max_bytes_for_level_base=2097152 --target_file_size_base=2097152 --periodic_compaction_seconds=0 --use_put_entity_one_in=10 --use_get_entity=1 --duration=60 --inter
val=10`
Reviewed By: ltamasi
Differential Revision: D49037040
Pulled By: jaykorean
fbshipit-source-id: a0648253ded6e91af7953de364ed3c6bf163626b
Summary:
**Context/Summary:** `rocksdb.file.read.db.open.micros` is landed in 8.3 but not 8.6. It was included in the HISTORY due to an error.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11839
Test Plan: no code change; Will cherry pick this to 8.6 branch when landed.
Reviewed By: anand1976
Differential Revision: D49294250
Pulled By: hx235
fbshipit-source-id: b2ac10758a15eadd5c129d80e93e1c3d0aa569cb
Summary:
When executing ClipColumnFamily, if end_key is equal to largest_user_key in a file, this key will not be deleted. So we need to change less than to less than or equal to
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11811
Reviewed By: ajkr
Differential Revision: D49206936
Pulled By: cbi42
fbshipit-source-id: 3e8bcb7b52040a9b4d1176de727616cc298d3445
Summary:
`last_stats_dump_time_microsec_` is not used after initialization.
I guess that it was previously used to implement periodically dumping stats,
but this functionality has now been delegated to the `PeriodicTaskScheduler`.
4b79e8c003/db/db_impl/db_impl.cc (L770-L778)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11824
Reviewed By: cbi42
Differential Revision: D49278311
Pulled By: jowlyzhang
fbshipit-source-id: 5856245580afc026e6b490755a45c5436a2375c9
Summary:
As discussed in https://github.com/facebook/rocksdb/issues/11730 , this PR tracks the effective `full_history_ts_low` per SuperVersion and update existing sanity checks for `ReadOptions.timestamp >= full_history_ts_low` to use this per SuperVersion `full_history_ts_low` instead. This also means the check is moved to happen after acquiring SuperVersion.
There are two motivations for this: 1) Each time `full_history_ts_low` really come into effect to collapse history, a new SuperVersion is always installed, because it would involve either a Flush or Compaction, both of which change the LSM tree shape. We can take advantage of this to ensure that as long as this sanity check is passed, even if `full_history_ts_low` can be concurrently increased and collapse some history above the requested `ReadOptions.timestamp`, a read request won’t have visibility to that part of history through this SuperVersion that it already acquired. 2) the existing sanity check uses `ColumnFamilyData::GetFullHistoryTsLow` without locking the db mutex, which is the mutex all `IncreaseFullHistoryTsLow` operation is using when mutating this field. So there is a race condition. This also solve the race condition on the read path.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11784
Test Plan:
`make all check`
// Checks success scenario really provide the read consistency attribute as mentioned above.
`./db_with_timestamp_basic_test --gtest_filter=*FullHistoryTsLowSanityCheckPassReadIsConsistent*`
// Checks failure scenario cleans up SuperVersion properly.
`./db_with_timestamp_basic_test --gtest_filter=*FullHistoryTsLowSanityCheckFail*`
`./db_secondary_test --gtest_filter=*FullHistoryTsLowSanityCheckFail*`
`./db_readonly_with_timestamp_test --gtest_filter=*FullHistoryTsLowSanitchCheckFail*`
Reviewed By: ltamasi
Differential Revision: D48894795
Pulled By: jowlyzhang
fbshipit-source-id: 1f801fe8e1bc8e63ca76c03cbdbd0974e5ff5bf6
Summary:
**Context/Summary:**
A size amp compaction can select and prevent a large number of L0 files from being selected by other compaction. If such compaction is running long or being queued behind, these L0 files will exist for long. With a few more flushes, we can run into write stop triggered by # L0 files. We've seen this happen on a host with many DBs sharing same thread pool, each of these DBs submits a size amp compaction with (110-180)+ files to the pool upon reopen and with a few more flushes, they hit the 200 L0 write stop condition.
The idea is to exclude some L0 input files in size amp compaction that are harmless to size amp reduction but improve the situation described above.
The exclusion algorithm is in `MightExcludeNewL0sToReduceWriteStop()` with two elements:
1. #L0 to exclude + (level0_stop_writes_trigger - num_l0_input_pre_exclusion) should be in the range of [min_merge_width, max_merge_width].
- This is to ensure we are excluding enough L0 input files but not too many to be qualified to picked for another compaction along with the incoming future L0 files before write stop.
2. Based on (1), further constrain #L0 to exclude based on the post-exclusion compaction score. The goal is to ensure our exclusion will not disqualify the size amp compaction from being a size amp compaction after exclusion.
**Tets plan:** New unit test
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11749
Reviewed By: ajkr
Differential Revision: D48850631
Pulled By: hx235
fbshipit-source-id: 2c321036e164087c36319dd5645cbbf6b6152092
Summary:
Existing compaction statistics are `COMPACTION_TIME` and `COMPACTION_CPU_TIME` which are histogram and are logged at the end of a compaction. The new statistics `COMPACTION_CPU_TOTAL_TIME` is for cumulative total compaction time which is updated regularly during a compaction. This allows user to more closely track compaction cpu usage.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11741
Test Plan: * new unit test `DBTestWithParam.CompactionTotalTimeTest`
Reviewed By: ajkr
Differential Revision: D48608094
Pulled By: cbi42
fbshipit-source-id: b597109f3e4bf2237fb5a216b6fd036e5363b4c0
Summary:
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11823
Similarly to https://github.com/facebook/rocksdb/pull/11813, the patch is a small refactoring that eliminates some copy-paste around sorting the columns of entities by column name.
Reviewed By: jaykorean
Differential Revision: D49195504
fbshipit-source-id: d48c9f290e3203f838cc5949856c469ecf730008
Summary:
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11807
For now, RocksDB has limited support for using merge with wide columns: when a bunch of merge operands have to be applied to a wide-column base value, RocksDB currently passes only the value of the default column to the application's `MergeOperator`, which means there is no way to update any other columns during a merge. As a first step in making this more general, the patch adds a new API `FullMergeV3` to `MergeOperator`.
`FullMergeV3`'s interface enables applications to receive a plain, wide-column, or non-existent base value as merge input, and to produce a new plain value, a new wide-column value, or an existing operand as merge output. Note that there are no limitations on the column names and values if the merge result is a wide-column entity. Also, the interface is general in the sense that it makes it possible e.g. for a merge that takes a plain base value and some deltas to produce a wide-column entity as a result.
For backward compatibility, the default implementation of `FullMergeV3` falls back to `FullMergeV2` and implements the current logic where merge operands are applied to the default column of the base entity and any other columns are unchanged. (Note that with `FullMergeV3` in the `MergeOperator` interface, this behavior will become customizable.)
This patch just introduces the new API and the default backward compatible implementation. I plan to integrate `FullMergeV3` into the query and compaction logic in subsequent diffs.
Reviewed By: jaykorean
Differential Revision: D49117253
fbshipit-source-id: 109e016f25cd130fc504790818d927bae7fec6bd
Summary:
During Seek, the iterator seeks every file on L0. In async_io, it submit the requests to seek on every file on L0 asynchronously using RocksDB FilePrefetchBuffer.
However, FilePrefetchBuffer does alignment and reads extra bytes then needed that can increase the throughput.
In case of non direct io, the alignment can be avoided.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11793
Test Plan:
- Added a unit test that fails without this PR.
- make crash_test -j32 completed successfully
Reviewed By: anand1976
Differential Revision: D48985051
Pulled By: akankshamahajan15
fbshipit-source-id: 2d130a9e7c3df9c4fcd0408406e6277ab75a4389
Summary:
This function uses racing reads for heuristic performance improvement. My change in https://github.com/facebook/rocksdb/issues/11792 only worked for clang, not gcc, and gcc does not accurately handle TSAN suppressions. I would have to mark much more code as suppressed than I want to.
So I've taken a different approach: TSAN build does not use the racing reads but substitutes random results, as an extra test that a "correct" value is not needed for correct overall behavior.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11806
Test Plan: manual TSAN builds & tests with cache_bench
Reviewed By: ajkr
Differential Revision: D49100115
Pulled By: pdillinger
fbshipit-source-id: d6d0dfb796d710b953212dd3fc171b6e88fadea1
Summary:
This is for fb internal use. Please see the comment in internal Phabricator for details.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11803
Reviewed By: hx235
Differential Revision: D49065093
Pulled By: jaykorean
fbshipit-source-id: acd71d7c1163f3c95c59c427caf944dacfe58ef6
Summary:
Tests a scenario where range tombstone reseek used to cause MergingIterator to discard non-ok status.
Ran on main without https://github.com/facebook/rocksdb/issues/11786:
```
./db_range_del_test --gtest_filter="*RangeDelReseekAfterFileReadError*"
Note: Google Test filter = *RangeDelReseekAfterFileReadError*
[==========] Running 1 test from 1 test case.
[----------] Global test environment set-up.
[----------] 1 test from DBRangeDelTest
[ RUN ] DBRangeDelTest.RangeDelReseekAfterFileReadError
db/db_range_del_test.cc:3577: Failure
Value of: iter->Valid()
Actual: true
Expected: false
[ FAILED ] DBRangeDelTest.RangeDelReseekAfterFileReadError (64 ms)
```
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11790
Reviewed By: ajkr
Differential Revision: D48972869
Pulled By: cbi42
fbshipit-source-id: b1a71867533b0fb60af86f8ce8a9e391ba84dd57
Summary:
https://github.com/facebook/rocksdb/issues/11789 added error injection during compaction to db_stress. However, error injection was not disabled after compaction completion, which resulted in some test failures due to stale errors.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11798
Reviewed By: cbi42
Differential Revision: D49022821
Pulled By: anand1976
fbshipit-source-id: 3cbfe18d55bee393697e063d05e7a7a7f88b7635
Summary:
Some repro unit tests for the bug fixed in https://github.com/facebook/rocksdb/pull/11782.
Ran on main without https://github.com/facebook/rocksdb/pull/11782:
```
./db_compaction_test --gtest_filter='*ErrorWhenReadFileHead'
Note: Google Test filter = *ErrorWhenReadFileHead
[==========] Running 1 test from 1 test case.
[----------] Global test environment set-up.
[----------] 1 test from DBCompactionTest
[ RUN ] DBCompactionTest.ErrorWhenReadFileHead
db/db_compaction_test.cc:10105: Failure
Value of: s.IsIOError()
Actual: false
Expected: true
[ FAILED ] DBCompactionTest.ErrorWhenReadFileHead (3960 ms)
./db_iterator_test --gtest_filter="*ErrorWhenReadFile*"
Note: Google Test filter = *ErrorWhenReadFile*
[==========] Running 1 test from 1 test case.
[----------] Global test environment set-up.
[----------] 1 test from DBIteratorTest
[ RUN ] DBIteratorTest.ErrorWhenReadFile
db/db_iterator_test.cc:3399: Failure
Value of: (iter->status()).ok()
Actual: true
Expected: false
[ FAILED ] DBIteratorTest.ErrorWhenReadFile (280 ms)
[----------] 1 test from DBIteratorTest (280 ms total)
```
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11788
Reviewed By: ajkr
Differential Revision: D48940284
Pulled By: cbi42
fbshipit-source-id: 06f3c5963f576db3f85d305ffb2745ee13d209bb
Summary:
Currently, rocksdb users would use the event listener to catch the compaction/flush event and log them if any. But now the reason is an integer type instead of a human-readable string, so we would like to convert them into a human-readable string.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11778
Reviewed By: jaykorean
Differential Revision: D49012934
Pulled By: ajkr
fbshipit-source-id: a4935b95d70c1be02aec65da7bf1c98a8cf8b933
Summary:
There was a `#include "port/lang.h"` situated inside an `extern "C" {` which just started causing the header to be unusuable in some contexts. This was a regression on the CircleCI job build-linux-unity-and-headers in https://github.com/facebook/rocksdb/issues/11792
The include, and another like it, now appears obsolete so removed.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11797
Test Plan: local `make check-headers` and `make`, CI
Reviewed By: jaykorean
Differential Revision: D48976826
Pulled By: pdillinger
fbshipit-source-id: 131d66969e045f2ded0f8936924ee30c9ef2655a
Summary:
- Fixed misspellings of "inject"
- Made user read errors retryable when `FLAGS_inject_error_severity == 1`
- Added compaction read errors when `FLAGS_read_fault_one_in > 0`. These are always retryable so that the DB will keep accepting writes
- Reenabled setting `compaction_readahead_size` in crash test. The reason for disabling it was to "keep the test clean", which is not a good enough reason to skip testing it
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11789
Test Plan:
With https://github.com/facebook/rocksdb/issues/11782 reverted, reproduced the bug:
- Build: `make -j56 db_stress`
- Command: `TEST_TMPDIR=/dev/shm python3 tools/db_crashtest.py blackbox --simple --write_buffer_size=524288 --target_file_size_base=524288 --max_bytes_for_level_base=2097152 --interval=10 --max_key=1000000`
- Output:
```
stderr has error message:
***put or merge error: Corruption: Compaction number of input keys does not match number of keys processed.***
```
Reviewed By: cbi42
Differential Revision: D48939994
Pulled By: ajkr
fbshipit-source-id: a1efb799efecdfd5d9cfd185e4a6321db8fccfbb
Summary:
[Apache Kvrocks](https://github.com/apache/kvrocks) is an open-source distributed key-value NoSQL database built on top of RocksDB. It serves as a cost-saving and capacity-increasing alternative drop-in replacement for Redis
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11779
Reviewed By: ajkr
Differential Revision: D48872257
Pulled By: cbi42
fbshipit-source-id: 507f67d69b826607a1464a22ec7c60abe11c5124
Summary:
This change add an experimental next-generation HyperClockCache (HCC) with automatic sizing of the underlying hash table. Both the existing version (stable) and the new version (experimental for now) of HCC are available depending on whether an estimated average entry charge is provided in HyperClockCacheOptions.
Internally, we call the two implementations AutoHyperClockCache (new) and FixedHyperClockCache (existing). The performance characteristics and much of the underlying logic are similar enough that AutoHCC is likely to make FixedHCC obsolete, and so it's best considered an evolution of the same technology or solution rather than an alternative. More specifically, both implementations share essentially the same logic for managing the state of individual entries in the cache, including metadata for reference counting and counting clocks for eviction. This metadata, which I like to call the "low-level HCC protocol," includes a read-write lock on entries, but relaxed consistency requirements on the cache (e.g. allowing rare duplication) means high-level cache operations never wait for these low-level per-entry locks. FixedHCC is fully wait-free.
AutoHCC is different in how entries are indexed into an efficient hash table. AutoHCC is "essentially wait-free" as there is no pattern of typical high-level operations on a large cache that can lead to one thread waiting on another to complete some work, though it can happen in some unusual/unlucky cases, or atypical uses such as erasing specific cache keys. Table growth and entry reclamation is more complex in AutoHCC compared to FixedHCC, so uses some localized locking to manage that. AutoHCC uses linear hashing to grow the table as needed, with low latency and to a precise size. AutoHCC depends on anonymous mmap support from the OS (currently verified working on Linux, MacOS, and Windows) to allow the array underlying a hash table to grow in place without wasting resident memory on space reserved but unused. AutoHCC uses a form of chaining while FixedHCC uses open addressing and double hashing.
More specifics:
* In developing this PR, a rare availability bug (minor) was noticed in the existing HCC implementation of Release()+erase_if_last_ref, which is now inherited into AutoHCC. Fixing this without a performance regression will not be simple, so is left for follow-up work.
* Some existing unit tests required adjustment of operational parameters or conditions to work with the new behaviors of AutoHCC. A number of bugs were found and fixed in the validation process, including getting unit tests in good working order.
* Added an option to cache_bench, `-degenerate_hash_bits` for correctness stress testing described below. For this, the tool uses the reverse-engineered hash function for HCC to generate keys in which the specified number of hash bits, in critical positions, have a fixed value. Essentially each degenerate hash bit will half the number of chain heads utilized and double the average chain length.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11738
Test Plan:
unit tests updated, and already added to db crash test. Also
## Correctness
The code includes generous assertions to check for unexpected states, especially at destruction time, so should be able to detect critical concurrency bugs. Less serious "availability bugs" in which cache data is hidden or cleanly lost are more difficult to detect, but also less scary for data correctness (as long as performance is good and the design is sound).
In average operation, the structure is extremely low stress and low contention (see next section) so stressing the corner case logic requires artificially stressing the operating conditions. First, we keep the structure small to increase the number of threads hitting the same chain or entry, and just one cache shard. Second, we artificially degrade the hashing so that chains are much longer than typical, using the new `-degenerate_hash_bits` option to cache_bench. Third, we re-create the structure from scratch frequently in order to exercise the Grow logic repeatedly and to get the benefit of the consistency checks in the structure's destructor in debug builds. For cache_bench this also means disabling the single-threaded "populate cache" step (normally used for steady state performance testing). And of course use many more threads than cores to have many preemptions.
An effective test for working out bugs was this (using debug build of course):
```
while ./cache_bench -cache_type=auto_hyper_clock_cache -histograms=0 -cache_size=8000000 -threads=100 -populate_cache=0 -ops_per_thread=10000 -degenerate_hash_bits=6 -num_shard_bits=0; do :; done
```
Or even smaller cases. This setup has around 27 utilized chains, with around 35 entries each, and yield-waits more than 1 million times per second (very high contention; see next section). I have let this run for hours searching for any lingering issues.
I've also run cache_bench under ASAN, UBSAN, and TSAN.
## Essentially wait free
There is a counter for number of yield() calls when one thread is waiting on another. When we pre-populate the structure in a single thread,
```
./cache_bench -cache_type=auto_hyper_clock_cache -histograms=0 -populate_cache=1 -ops_per_thread=200000 2>&1 | grep Yield
```
We see something on the order of 1 yield call per second across 16 threads, even when we load the system other other jobs (parallel compilation). With -populate_cache=0, there are more yield opportunities with parallel table growth. On an otherwise unloaded system, we still see very small (single digit) yield counts, with a chance of getting into the thousands, and getting into 10s of thousands per second during table growth phase if the system is loaded with other jobs. However, I am not worried about this if performance is still good (see next section).
## Overall performance
Although cache_bench initially suggested performance very close to FixedHCC, there was a very noticeable performance hit under a db_bench setup like used in validating https://github.com/facebook/rocksdb/issues/10626. Much of the difference has been reduced by optimizing Lookup with a "naive" pass that will almost always find entries quickly, and only falling back to the careful Lookup algorithm when not found in the first pass.
Setups (chosen to be sensitive to block cache performance), and compiled with USE_CLANG=1 JEMALLOC=1 PORTABLE=0 DEBUG_LEVEL=0:
```
TEST_TMPDIR=/dev/shm base/db_bench -benchmarks=fillrandom -num=30000000 -disable_wal=1 -bloom_bits=16
```
### No regression on FixedHCC
Running before & after builds at the same time on a 48 core machine.
```
TEST_TMPDIR=/dev/shm /usr/bin/time ./db_bench -benchmarks=readrandom[-X10],block_cache_entry_stats,cache_report_problems -readonly -num=30000000 -bloom_bits=16 -cache_index_and_filter_blocks=1 -cache_size=610000000 -duration 20 -threads=24 -cache_type=fixed_hyper_clock_cache -seed=1234
```
Before:
readrandom [AVG 10 runs] : 847234 (± 8150) ops/sec; 59.2 (± 0.6) MB/sec
703MB max RSS
After:
readrandom [AVG 10 runs] : 851021 (± 7929) ops/sec; 59.5 (± 0.6) MB/sec
706MB max RSS
Probably no material difference.
### Single-threaded performance
Using `[-X2]` and `-threads=1` and `-duration=30`, running all three at the same time:
lru_cache: 55100 ops/sec, then 55862 ops/sec (627MB max RSS)
fixed_hyper_clock_cache: 60496 ops/sec, then 61231 ops/sec (626MB max RSS)
auto_hyper_clock_cache: 47560 ops/sec, then 56081 ops/sec (626MB max RSS)
So AutoHCC has more ramp-up cost in the first pass as the cache grows to the appropriate size. (In single-threaded operation, the parallelizability and per-op low latency of table growth is overall slower.) However, once up to size, its performance is comparable to LRUCache. FixedHCC's lean operations still win overall when a good estimate is available.
If we look at HCC table stats, we can see that this configuration is not favorable to AutoHCC (and I have verified that other memory sizes do not yield substantially different results, until shards are under-sized for the full filters):
FixedHCC:
Slot occupancy stats: Overall 47% (124991/262144), Min/Max/Window = 28%/64%/500, MaxRun{Pos/Neg} = 17/22
AutoHCC:
Slot occupancy stats: Overall 59% (125781/209682), Min/Max/Window = 43%/82%/500, MaxRun{Pos/Neg} = 76/16
Head occupancy stats: Overall 43% (92259/209682), Min/Max/Window = 24%/74%/500, MaxRun{Pos/Neg} = 19/26
Entries at home count: 53350
FixedHCC configuration is relatively good for speed, and not ideal for space utilization. As is typical, AutoHCC has tighter control on metadata usage (209682 x 64 bytes rather than 262144 x 64 bytes), and the higher load factor is slightly worse for speed. LRUCache also has more metadata usage, at 199680 x 96 bytes of tracked metadata (plus roughly another 10% of that untracked in the head pointers), and that metadata is subject to fragmentation.
### Parallel performance, high hit rate
Now using `[-X10]` and `-threads=10`, all three at the same time
lru_cache: [AVG 10 runs] : 263629 (± 1425) ops/sec; 18.4 (± 0.1) MB/sec
655MB max RSS, 97.1% cache hit rate
fixed_hyper_clock_cache: [AVG 10 runs] : 479590 (± 8114) ops/sec; 33.5 (± 0.6) MB/sec
651MB max RSS, 97.1% cache hit rate
auto_hyper_clock_cache: [AVG 10 runs] : 418687 (± 5915) ops/sec; 29.3 (± 0.4) MB/sec
657MB max RSS, 97.1% cache hit rate
Even with just 10-way parallelism for each cache (though 30+/48 cores busy overall), LRUCache is already showing performance degradation, while AutoHCC is in the neighborhood of FixedHCC. And that brings us to the question of how AutoHCC holds up under extreme parallelism, so now independent runs with `-threads=100` (overloading 48 cores).
lru_cache: 438613 ops/sec, 827MB max RSS
fixed_hyper_clock_cache: 1651310 ops/sec, 812MB max RSS
auto_hyper_clock_cache: 1505875 ops/sec, 821MB max RSS (Yield count: 1089 over 30s)
Clearly, AutoHCC holds up extremely well under extreme parallelism, even closing some of the modest performance gap with FixedHCC.
### Parallel performance, low hit rate
To get down to roughly 50% cache hit rate, we use `-cache_index_and_filter_blocks=0 -cache_size=1650000000` with `-threads=10`. Here the extra cost of running counting clock eviction, especially on the chains of AutoHCC, are evident, especially with the lower contention of cache_index_and_filter_blocks=0:
lru_cache: 725231 ops/sec, 1770MB max RSS, 51.3% hit rate
fixed_hyper_clock_cache: 638620 ops/sec, 1765MB max RSS, 50.2% hit rate
auto_hyper_clock_cache: 541018 ops/sec, 1777MB max RSS, 50.8% hit rate
Reviewed By: jowlyzhang
Differential Revision: D48784755
Pulled By: pdillinger
fbshipit-source-id: e79813dc087474ac427637dd282a14fa3011a6e4
Summary:
This should only affect iterator when
- user uses DeleteRange(),
- An iterator from level L has a non-ok status (such non-ok status may not be caught before the bug fix in https://github.com/facebook/rocksdb/pull/11783), and
- A range tombstone covers a key from level > L and triggers a reseek sets the status_ to OK in SeekImpl()/SeekPrevImpl() e.g. bd6a8340c3/table/merging_iterator.cc (L801)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11786
Differential Revision: D48908830
Pulled By: cbi42
fbshipit-source-id: eb564be375af4e33dc27542eff753260186e6d5d
Summary:
This happens in (Compaction)MergingIterator layer, and can cause data loss during compaction or read/scan return incorrect result
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11782
Reviewed By: ajkr
Differential Revision: D48880575
Pulled By: cbi42
fbshipit-source-id: 2294ad284a6d653d3674bebe55380f12ee4b645b
Summary:
**Context/Summary:**
After https://github.com/facebook/rocksdb/pull/11631, we rely on `compaction_readahead_size` for how much to read ahead for compaction read under non-direct IO case. https://github.com/facebook/rocksdb/pull/11658 therefore also sanitized 0 `compaction_readahead_size` to 2MB under non-direct IO, which is consistent with the existing sanitization with direct IO.
However, this makes disabling compaction readahead impossible as well as add one more scenario to the inconsistent effects between `Options.compaction_readahead_size=0` during DB open and `SetDBOptions("compaction_readahead_size", "0")` .
- `SetDBOptions("compaction_readahead_size", "0")` will disable compaction readahead as its logic never goes through sanitization above while `Options.compaction_readahead_size=0` will go through sanitization.
Therefore we decided to do this PR.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11762
Test Plan: Modified existing UTs to cover this PR
Reviewed By: ajkr
Differential Revision: D48759560
Pulled By: hx235
fbshipit-source-id: b3f85e58bda362a6fa1dc26bd8a87aa0e171af79
Summary:
For a SST file that uses user-defined timestamp aware comparators, if a lower or upper bound is set, sst_dump tool doesn't handle it well. This PR adds support for that. While working on this `MaybeAddTimestampsToRange` is moved to the udt_util.h file to be shared.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11757
Test Plan:
make all check
for changes in db_impl.cc and db_impl_compaction_flush.cc
for changes in sst_file_dumper.cc, I manually tested this change handles specifying bounds for UDT use cases. It probably should have a unit test file eventually.
Reviewed By: ltamasi
Differential Revision: D48668048
Pulled By: jowlyzhang
fbshipit-source-id: 1560465f40e44668d6d82a7439fe9012be0e74a8
Summary:
wide_columns can now be pretty-printed in the following commands
- `./ldb dump_wal`
- `./ldb dump`
- `./ldb idump`
- `./ldb dump_live_files`
- `./ldb scan`
- `./sst_dump --command=scan`
There are opportunities to refactor to reduce some nearly identical code. This PR is initial change to add wide column support in `ldb` and `sst_dump` tool. More PRs to come for the refactor.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11754
Test Plan:
**New Tests added**
- `WideColumnsHelperTest::DumpWideColumns`
- `WideColumnsHelperTest::DumpSliceAsWideColumns`
**Changes added to existing tests**
- `ExternalSSTFileTest::BasicMixed` added to cover mixed case (This test should have been added in https://github.com/facebook/rocksdb/issues/11688). This test does not verify the ldb or sst_dump output. This test was used to create test SST files having some rows with wide columns and some without and the generated SST files were used to manually test sst_dump_tool.
- `createSST()` in `sst_dump_test` now takes `wide_column_one_in` to add wide column value in SST
**dump_wal**
```
./ldb dump_wal --walfile=/tmp/rocksdbtest-226125/db_wide_basic_test_2675429_2308393776696827948/000004.log --print_value --header
```
```
Sequence,Count,ByteSize,Physical Offset,Key(s) : value
1,1,59,0,PUT_ENTITY(0) : 0x:0x68656C6C6F 0x617474725F6E616D6531:0x666F6F 0x617474725F6E616D6532:0x626172
2,1,34,42,PUT_ENTITY(0) : 0x617474725F6F6E65:0x74776F 0x617474725F7468726565:0x666F7572
3,1,17,7d,PUT(0) : 0x7468697264 : 0x62617A
```
**idump**
```
./ldb --db=/tmp/rocksdbtest-226125/db_wide_basic_test_3481961_2308393776696827948/ idump
```
```
'first' seq:1, type:22 => :hello attr_name1:foo attr_name2:bar
'second' seq:2, type:22 => attr_one:two attr_three:four
'third' seq:3, type:1 => baz
Internal keys in range: 3
```
**SST Dump from dump_live_files**
```
./ldb --db=/tmp/rocksdbtest-226125/db_wide_basic_test_3481961_2308393776696827948/ compact
./ldb --db=/tmp/rocksdbtest-226125/db_wide_basic_test_3481961_2308393776696827948/ dump_live_files
```
```
...
==============================
SST Files
==============================
/tmp/rocksdbtest-226125/db_wide_basic_test_3481961_2308393776696827948/000013.sst level:1
------------------------------
Process /tmp/rocksdbtest-226125/db_wide_basic_test_3481961_2308393776696827948/000013.sst
Sst file format: block-based
'first' seq:0, type:22 => :hello attr_name1:foo attr_name2:bar
'second' seq:0, type:22 => attr_one:two attr_three:four
'third' seq:0, type:1 => baz
...
```
**dump**
```
./ldb --db=/tmp/rocksdbtest-226125/db_wide_basic_test_3481961_2308393776696827948/ dump
```
```
first ==> :hello attr_name1:foo attr_name2:bar
second ==> attr_one:two attr_three:four
third ==> baz
Keys in range: 3
```
**scan**
```
./ldb --db=/tmp/rocksdbtest-226125/db_wide_basic_test_3481961_2308393776696827948/ scan
```
```
first : :hello attr_name1:foo attr_name2:bar
second : attr_one:two attr_three:four
third : baz
```
**sst_dump**
```
./sst_dump --file=/tmp/rocksdbtest-226125/db_wide_basic_test_3481961_2308393776696827948/000013.sst --command=scan
```
```
options.env is 0x7ff54b296000
Process /tmp/rocksdbtest-226125/db_wide_basic_test_3481961_2308393776696827948/000013.sst
Sst file format: block-based
from [] to []
'first' seq:0, type:22 => :hello attr_name1:foo attr_name2:bar
'second' seq:0, type:22 => attr_one:two attr_three:four
'third' seq:0, type:1 => baz
```
Reviewed By: ltamasi
Differential Revision: D48837999
Pulled By: jaykorean
fbshipit-source-id: b0280f0589d2b9716bb9b50530ffcabb397d140f
Summary:
This PR adds a missing piece for the UDT in memtable only feature, which is to automatically increase `full_history_ts_low` when flush happens during recovery.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11774
Test Plan:
Added unit test
make all check
Reviewed By: ltamasi
Differential Revision: D48799109
Pulled By: jowlyzhang
fbshipit-source-id: fd681ed66d9d40904ca2c919b2618eb692686035
