Andrew Kryczka [Mon, 18 Jul 2022 21:48:42 +0000 (14:48 -0700)]
Make RateLimiter not Customizable (#10378)
Summary:
(PR created for informational/testing purposes only.)
- Fixes lost dynamic updates to GenericRateLimiter bandwidth using `SetBytesPerSecond()`
- Benefit over #10374 is eliminating race conditions with Configurable framework.
Yanqin Jin [Wed, 13 Jul 2022 00:16:57 +0000 (17:16 -0700)]
Stop tracking syncing live WAL for performance (#10330)
Summary:
With https://github.com/facebook/rocksdb/issues/10087, applications calling `SyncWAL()` or writing with `WriteOptions::sync=true` can suffer
from performance regression. This PR reverts to original behavior of tracking the syncing of closed WALs.
After we revert back to old behavior, recovery, whether kPointInTime or kAbsoluteConsistency, may fail to
detect corruption in synced WALs if the corruption is in the live WAL.
Fix bug in Logger creation if dbname and db_log_dir are on different filesystem (#10292)
Summary:
If dbname and db_log_dir are at different filesystems (one
local and one remote), creation of dbname will fail because that path
doesn't exist wrt to db_log_dir.
This patch will ignore the error returned on creation of dbname. If they
are on same filesystem, db_log_dir creation will automatically return
the error in case there is any error in creation of dbname.
Bo Wang [Fri, 24 Jun 2022 23:13:44 +0000 (16:13 -0700)]
Pass rate_limiter_priority through filter block reader functions to FS (#10251)
Summary:
With https://github.com/facebook/rocksdb/pull/9996 , we can pass the rate_limiter_priority to FS for most cases. This PR is to update the code path for filter block reader.
Yanqin Jin [Wed, 22 Jun 2022 15:26:38 +0000 (08:26 -0700)]
Expose the initial logger creation error (#10223)
Summary:
https://github.com/facebook/rocksdb/issues/9984 changes the behavior of RocksDB: if logger creation failed during `SanitizeOptions()`,
`DB::Open()` will fail. However, since `SanitizeOptions()` is called in `DBImpl::DBImpl()`, we cannot
directly expose the error to caller without some additional work.
This is a first version proposal which:
- Adds a new member `init_logger_creation_s` to `DBImpl` to store the result of init logger creation
- Checks the error during `DB::Open()` and return it to caller if non-ok
This is not very ideal. We can alternatively move the logger creation logic out of the `SanitizeOptions()`.
Since `SanitizeOptions()` is used in other places, we need to check whether this change breaks anything
in case other callers of `SanitizeOptions()` assumes that a logger should be created.
Peter Dillinger [Tue, 21 Jun 2022 23:23:58 +0000 (16:23 -0700)]
Add data block hash index to crash test, fix MultiGet issue (#10220)
Summary:
There was a bug in the MultiGet enhancement in https://github.com/facebook/rocksdb/issues/9899 with data
block hash index, which was not caught because data block hash index was
never added to stress tests. This change fixes both issues.
Test Plan:
Failure quickly reproduces in crash test with
kDataBlockBinaryAndHash, and does not seem to with the fix. Reproducing
the failure with a unit test I believe would be too tricky and fragile
to be worthwhile.
Peter Dillinger [Tue, 21 Jun 2022 00:42:01 +0000 (17:42 -0700)]
Fix bad include (#10213)
Summary:
include "include/rocksdb/blah.h" is messing up some internal
builds vs. include "rocksdb/blah." This fixes the bad case and adds a
check for future instances.
Changyu Bi [Sat, 18 Jun 2022 22:12:17 +0000 (15:12 -0700)]
Fix a bug in WriteBatchInternal::Append when write batch KV protection is turned on (#10201)
Summary:
This bug was discovered after write batch checksum verification before WAL is added (https://github.com/facebook/rocksdb/issues/10114) and stress test with write batch checksum protection is turned on (https://github.com/facebook/rocksdb/issues/10037). In this [line](https://github.com/facebook/rocksdb/blob/d5d8920f2cfd06d1803b0976acbe8b564b88b6b1/db/write_batch.cc#L2887), the number of checksums may not be consistent with `batch->Count()`. This PR fixes this issue.
Andrew Kryczka [Fri, 17 Jun 2022 23:45:28 +0000 (16:45 -0700)]
Fix race condition with WAL tracking and `FlushWAL(true /* sync */)` (#10185)
Summary:
`FlushWAL(true /* sync */)` is used internally and for manual WAL sync. It had a bug when used together with `track_and_verify_wals_in_manifest` where the synced size tracked in MANIFEST was larger than the number of bytes actually synced.
The bug could be repro'd almost immediately with the following crash test command: `python3 tools/db_crashtest.py blackbox --simple --write_buffer_size=524288 --max_bytes_for_level_base=2097152 --target_file_size_base=524288 --duration=3600 --interval=10 --sync_fault_injection=1 --disable_wal=0 --checkpoint_one_in=1000 --max_key=10000 --value_size_mult=33`.
An example error message produced by the above command is shown below. The error sometimes arose from the checkpoint and other times arose from the main stress test DB.
```
Corruption: Size mismatch: WAL (log number: 119) in MANIFEST is 27938 bytes , but actually is 27859 bytes on disk.
```
Test Plan:
- repro unit test
- the above crash test command no longer finds the error. It does find a different error after a while longer such as "Corruption: WAL file 481 required by manifest but not in directory list"
Hui Xiao [Fri, 17 Jun 2022 23:40:47 +0000 (16:40 -0700)]
Add rate-limiting support to batched MultiGet() (#10159)
Summary:
**Context/Summary:**
https://github.com/facebook/rocksdb/pull/9424 added rate-limiting support for user reads, which does not include batched `MultiGet()`s that call `RandomAccessFileReader::MultiRead()`. The reason is that it's harder (compared with RandomAccessFileReader::Read()) to implement the ideal rate-limiting where we first call `RateLimiter::RequestToken()` for allowed bytes to multi-read and then consume those bytes by satisfying as many requests in `MultiRead()` as possible. For example, it can be tricky to decide whether we want partially fulfilled requests within one `MultiRead()` or not.
However, due to a recent urgent user request, we decide to pursue an elementary (but a conditionally ineffective) solution where we accumulate enough rate limiter requests toward the total bytes needed by one `MultiRead()` before doing that `MultiRead()`. This is not ideal when the total bytes are huge as we will actually consume a huge bandwidth from rate-limiter causing a burst on disk. This is not what we ultimately want with rate limiter. Therefore a follow-up work is noted through TODO comments.
Test Plan:
- Modified existing unit test `DBRateLimiterOnReadTest/DBRateLimiterOnReadTest.NewMultiGet`
- Traced the underlying system calls `io_uring_enter` and verified they are 10 seconds apart from each other correctly under the setting of `strace -ftt -e trace=io_uring_enter ./db_bench -benchmarks=multireadrandom -db=/dev/shm/testdb2 -readonly -num=50 -threads=1 -multiread_batched=1 -batch_size=100 -duration=10 -rate_limiter_bytes_per_sec=200 -rate_limiter_refill_period_us=1000000 -rate_limit_bg_reads=1 -disable_auto_compactions=1 -rate_limit_user_ops=1` where each `MultiRead()` read about 2000 bytes (inspected by debugger) and the rate limiter grants 200 bytes per seconds.
- Stress test:
- Verified `./db_stress (-test_cf_consistency=1/test_batches_snapshots=1) -use_multiget=1 -cache_size=1048576 -rate_limiter_bytes_per_sec=10241024 -rate_limit_bg_reads=1 -rate_limit_user_ops=1` work
Gang Liao [Fri, 17 Jun 2022 22:22:59 +0000 (15:22 -0700)]
Read blob from blob cache if exists when GetBlob() (#10178)
Summary:
There is currently no caching mechanism for blobs, which is not ideal especially when the database resides on remote storage (where we cannot rely on the OS page cache). As part of this task, we would like to make it possible for the application to configure a blob cache.
In this task, we added a new abstraction layer `BlobSource` to retrieve blobs from either blob cache or raw blob file. Note: For simplicity, the current PR only includes `GetBlob()`. `MultiGetBlob()` will be included in the next PR.
This PR is a part of https://github.com/facebook/rocksdb/issues/10156
Peter Dillinger [Fri, 17 Jun 2022 20:08:45 +0000 (13:08 -0700)]
Use optimized folly DistributedMutex in LRUCache when available (#10179)
Summary:
folly DistributedMutex is faster than standard mutexes though
imposes some static obligations on usage. See
https://github.com/facebook/folly/blob/main/folly/synchronization/DistributedMutex.h
for details. Here we use this alternative for our Cache implementations
(especially LRUCache) for better locking performance, when RocksDB is
compiled with folly.
Also added information about which distributed mutex implementation is
being used to cache_bench output and to DB LOG.
Intended follow-up:
* Use DMutex in more places, perhaps improving API to support non-scoped
locking
* Fix linking with fbcode compiler (needs ROCKSDB_NO_FBCODE=1 currently)
Credit: Thanks Siying for reminding me about this line of work that was previously
left unfinished.
Summary:
Added an option, `WriteOptions::protection_bytes_per_key`, that controls how many bytes per key we use for integrity protection in `WriteBatch`. It takes effect when `WriteBatch::GetProtectionBytesPerKey() == 0`.
Currently the only supported value is eight. Invoking a user API with it set to any other nonzero value will result in `Status::NotSupported` returned to the user.
There is also a bug fix for integrity protection with `inplace_callback`, where we forgot to take into account the possible change in varint length when calculating KV checksum for the final encoded buffer.
Test Plan:
- Manual
- Set default value of `WriteOptions::protection_bytes_per_key` to eight and ran `make check -j24`
- Enabled in MyShadow for 1+ week
- Automated
- Unit tests have a `WriteMode` that enables the integrity protection via `WriteOptions`
- Crash test - in most cases, use `WriteOptions::protection_bytes_per_key` to enable integrity protection
Peter Dillinger [Fri, 17 Jun 2022 04:14:10 +0000 (21:14 -0700)]
Fix a false negative merge conflict (#10192)
Summary:
.. between https://github.com/facebook/rocksdb/issues/10184 and https://github.com/facebook/rocksdb/issues/10122 not detected by source control,
leading to non-compiling code.
Peter Dillinger [Thu, 16 Jun 2022 23:41:25 +0000 (16:41 -0700)]
More testing w/prefix extractor, small refactor (#10122)
Summary:
There was an interesting code path not covered by testing that
is difficult to replicate in a unit test, which is now covered using a
sync point. Specifically, the case of table_prefix_extractor == null and
!need_upper_bound_check in `BlockBasedTable::PrefixMayMatch`, which
can happen if table reader is open before extractor is registered with global
object registry, but is later registered and re-set with SetOptions. (We
don't have sufficient testing control over object registry to set that up
repeatedly.)
Also, this function has been renamed to `PrefixRangeMayMatch` for clarity
vs. other functions that are not the same.
Peter Dillinger [Thu, 16 Jun 2022 22:51:33 +0000 (15:51 -0700)]
Remove deprecated block-based filter (#10184)
Summary:
In https://github.com/facebook/rocksdb/issues/9535, release 7.0, we hid the old block-based filter from being created using
the public API, because of its inefficiency. Although we normally maintain read compatibility
on old DBs forever, filters are not required for reading a DB, only for optimizing read
performance. Thus, it should be acceptable to remove this code and the substantial
maintenance burden it carries as useful features are developed and validated (such
as user timestamp).
This change completely removes the code for reading and writing the old block-based
filters, net removing about 1370 lines of code no longer needed. Options removed from
testing / benchmarking tools. The prior existence is only evident in a couple of places:
* `CacheEntryRole::kDeprecatedFilterBlock` - We can update this public API enum in
a major release to minimize source code incompatibilities.
* A warning is logged when an old table file is opened that used the old block-based
filter. This is provided as a courtesy, and would be a pain to unit test, so manual testing
should suffice. Unfortunately, sst_dump does not tell you whether a file uses
block-based filter, and the structure of the code makes it very difficult to fix.
* To detect that case, `kObsoleteFilterBlockPrefix` (renamed from `kFilterBlockPrefix`)
for metaindex is maintained (for now).
Other notes:
* In some cases where numbers are associated with filter configurations, we have had to
update the assigned numbers so that they all correspond to something that exists.
* Fixed potential stat counting bug by assuming `filter_checked = false` for cases
like `filter == nullptr` rather than assuming `filter_checked = true`
* Removed obsolete `block_offset` and `prefix_extractor` parameters from several
functions.
* Removed some unnecessary checks `if (!table_prefix_extractor() && !prefix_extractor)`
because the caller guarantees the prefix extractor exists and is compatible
anand76 [Thu, 16 Jun 2022 19:12:43 +0000 (12:12 -0700)]
Update stats to help users estimate MultiGet async IO impact (#10182)
Summary:
Add a couple of stats to help users estimate the impact of potential MultiGet perf improvements -
1. NUM_LEVEL_READ_PER_MULTIGET - A histogram stat for number of levels that required MultiGet to read from a file
2. MULTIGET_COROUTINE_COUNT - A ticker stat to count the number of times the coroutine version of MultiGetFromSST was used
The NUM_DATA_BLOCKS_READ_PER_LEVEL stat is obsoleted as it doesn't provide useful information for MultiGet optimization.
Yanqin Jin [Thu, 16 Jun 2022 05:00:24 +0000 (22:00 -0700)]
Abort in dbg mode after logging (#10183)
Summary:
In CompactionIterator code, there are multiple places where the process
will abort in dbg mode before logging the error message describing the
cause. This PR changes only the logging behavior for compaction iterator so
that error message is written to LOG before the process aborts in debug
mode.
Also updated the triggering condition for an assertion for single delete with
user-defined timestamp.
Akanksha Mahajan [Thu, 16 Jun 2022 03:17:35 +0000 (20:17 -0700)]
Add few optimizations in async_io for short scans (#10140)
Summary:
This PR adds few optimizations for async_io for shorter scans.
1. If async_io is enabled, seek would create FilePrefetchBuffer object to fetch the data asynchronously. However `FilePrefetchbuffer::num_file_reads_` wasn't taken into consideration if it calls Next after Seek and would go for Prefetching. This PR fixes that and Next will go for prefetching only if `FilePrefetchbuffer::num_file_reads_` is greater than 2 along with if blocks are sequential. This scenario is only for implicit auto readahead.
2. For seek, when it calls TryReadFromCacheAsync to poll it makes async call as well because TryReadFromCacheAsync flow wasn't changed. So I updated to return after poll instead of further prefetching any data.
Peter Dillinger [Wed, 15 Jun 2022 22:39:49 +0000 (15:39 -0700)]
Fix handling of accidental truncation of IDENTITY file (#10173)
Summary:
A consequence of https://github.com/facebook/rocksdb/issues/9990 was requiring a non-empty DB ID to generate
new SST files. But if the DB ID is not tracked in the manifest and the IDENTITY file
is somehow truncated to 0 bytes, then an empty DB ID would be assigned, leading
to crash. This change ensures a non-empty DB ID is assigned and set in the
IDENTITY file.
Also,
* Some light refactoring to clean up the logic
* (I/O efficiency) If the ID is tracked in the manifest and already matches the
IDENTITY file, don't needlessly overwrite the file.
* (Debugging) Log the DB ID to info log on open, because sometimes IDENTITY
can change if DB is moved around (though it would be unusual for info log to
be copied/moved without IDENTITY file)
Changyu Bi [Wed, 15 Jun 2022 20:43:58 +0000 (13:43 -0700)]
Verify write batch checksum before WAL (#10114)
Summary:
Context: WriteBatch can have key-value checksums when it was created `with protection_bytes_per_key > 0`.
This PR added checksum verification for write batches before they are written to WAL.
Ali Saidi [Wed, 15 Jun 2022 20:08:11 +0000 (13:08 -0700)]
Change the instruction used for a pause on arm64 (#10118)
Summary:
While the yield instruction conseptually sounds correct on most platforms it is
a simple nop that doesn't delay the execution anywhere close to what an x86
pause instruction does. In other projects with spin-wait loops an isb has been
observed to be much closer to the x86 behavior.
On a Graviton3 system the following test improves on average by 2x with this
change averaged over 20 runs:
sdong [Wed, 15 Jun 2022 20:05:58 +0000 (13:05 -0700)]
Use madvise() for mmaped file advise (#10170)
Summary:
A recent PR https://github.com/facebook/rocksdb/pull/10142 enabled fadvise for mmaped file. However, we were told that it might not take effective and madvise() should be used.
Yanqin Jin [Wed, 15 Jun 2022 19:38:04 +0000 (12:38 -0700)]
Allow db_bench and db_stress to set `allow_data_in_errors` (#10171)
Summary:
There is `Options::allow_data_in_errors` that controls whether RocksDB
is allowed to log data, e.g. key, value, etc in LOG files. It is false
by default. However, in db_bench and db_stress, it is often ok to log
data because there is no concern about privacy.
This PR allows db_stress and db_bench to set this option on the command
line, while it remains false by default. Furthermore, make
crash/recovery test driven by db_crashtest.py to opt-in.
iseki [Wed, 15 Jun 2022 04:32:10 +0000 (21:32 -0700)]
Fix C4702 on windows (#10146)
Summary:
This code is unreachable when `ROCKSDB_LITE` not defined. And it cause build fail on my environment VS2019 16.11.15.
```
-- Selecting Windows SDK version 10.0.19041.0 to target Windows 10.0.19044.
-- The CXX compiler identification is MSVC 19.29.30145.0
-- The C compiler identification is MSVC 19.29.30145.0
-- The ASM compiler identification is MSVC
```
mpoeter [Wed, 15 Jun 2022 04:29:52 +0000 (21:29 -0700)]
Fix potential leak when reusing PinnableSlice instances. (#10166)
Summary:
`PinnableSlice` may hold a handle to a cache value which must be released to correctly decrement the ref-counter. However, when `PinnableSlice` variables are reused, e.g. like this:
```
PinnableSlice pin_slice;
db.Get("foo", &pin_slice);
db.Get("foo", &pin_slice);
```
then the second `Get` simply overwrites the old value in `pin_slice` and the handle returned by the first `Get` is _not_ released.
This PR adds `Reset` calls to the `Get`/`MultiGet` calls that accept `PinnableSlice` arguments to ensure proper cleanup of old values.
Ali Saidi [Wed, 15 Jun 2022 00:58:44 +0000 (17:58 -0700)]
Modify the instructions emited for PREFETCH on arm64 (#10117)
Summary:
__builtin_prefetch(...., 1) prefetches into the L2 cache on x86 while the same
emits a pldl3keep instruction on arm64 which doesn't seem to be close enough.
Testing on a Graviton3, and M1 system with memtablerep_bench fillrandom and
skiplist througpuh increased as follows adjusting the 1 to 2 or 3:
```
1 -> 2 1 -> 3
----------------------------
Graviton3 +10% +15%
M1 +10% +10%
```
Given that prefetching into the L1 cache seems to help, I chose that conversion
Gang Liao [Tue, 14 Jun 2022 21:19:26 +0000 (14:19 -0700)]
Add blob cache option in the column family options (#10155)
Summary:
There is currently no caching mechanism for blobs, which is not ideal especially when the database resides on remote storage (where we cannot rely on the OS page cache). As part of this task, we would like to make it possible for the application to configure a blob cache.
This PR is a part of https://github.com/facebook/rocksdb/issues/10156
tabokie [Tue, 14 Jun 2022 20:20:54 +0000 (13:20 -0700)]
fix a false positive case of parsing table factory from options file (#10094)
Summary:
During options file parsing, reset table factory before attempting to parse it
from string. This avoids mistakenly treating the default table factory as a
newly created one.
Hui Xiao [Tue, 14 Jun 2022 20:06:40 +0000 (13:06 -0700)]
Account memory of FileMetaData in global memory limit (#9924)
Summary:
**Context/Summary:**
As revealed by heap profiling, allocation of `FileMetaData` for [newly created file added to a Version](https://github.com/facebook/rocksdb/pull/9924/files#diff-a6aa385940793f95a2c5b39cc670bd440c4547fa54fd44622f756382d5e47e43R774) can consume significant heap memory. This PR is to account that toward our global memory limit based on block cache capacity.
Test Plan:
- Previous `make check` verified there are only 2 places where the memory of the allocated `FileMetaData` can be released
- New unit test `TEST_P(ChargeFileMetadataTestWithParam, Basic)`
- db bench (CPU cost of `charge_file_metadata` in write and compact)
- **write micros/op: -0.24%** : `TEST_TMPDIR=/dev/shm/testdb ./db_bench -benchmarks=fillseq -db=$TEST_TMPDIR -charge_file_metadata=1 (remove this option for pre-PR) -disable_auto_compactions=1 -write_buffer_size=100000 -num=4000000 | egrep 'fillseq'`
- **compact micros/op -0.87%** : `TEST_TMPDIR=/dev/shm/testdb ./db_bench -benchmarks=fillseq -db=$TEST_TMPDIR -charge_file_metadata=1 -disable_auto_compactions=1 -write_buffer_size=100000 -num=4000000 -numdistinct=1000 && ./db_bench -benchmarks=compact -db=$TEST_TMPDIR -use_existing_db=1 -charge_file_metadata=1 -disable_auto_compactions=1 | egrep 'compact'`
Akanksha Mahajan [Tue, 14 Jun 2022 19:35:11 +0000 (12:35 -0700)]
Fix the failure related to io_uring_prep_cancel (#10165)
Summary:
Fix for Internal jobs are failing with
```
error: no matching function for call to 'io_uring_prep_cancel'
io_uring_prep_cancel(sqe, posix_handle, 0);
^~~~~~~~~~~~~~~~~~~~
note: candidate function not viable: no known conversion from 'rocksdb::Posix_IOHandle *' to '__u64' (aka 'unsigned long long') for 2nd argument
static inline void io_uring_prep_cancel(struct io_uring_sqe *sqe,
```
User data is set using `io_uring_set_data` API so no need to pass posix_handle here.
Make the per-shard hash table fixed-size. (#10154)
Summary:
We make the size of the per-shard hash table fixed. The base level of the hash table is now preallocated with the required capacity. The user must provide an estimate of the size of the values.
Notice that even though the base level becomes fixed, the chains are still dynamic. Overall, the shard capacity mechanisms haven't changed, so we don't need to test this.
Yanqin Jin [Tue, 14 Jun 2022 01:54:38 +0000 (18:54 -0700)]
Fix a race condition in transaction stress test (#10157)
Summary:
Before this PR, there can be a race condition between the thread calling
`StressTest::Open()` and a background compaction thread calling
`MultiOpsTxnsStressTest::VerifyPkSkFast()`.
```
Time thread1 bg_compact_thr
| TransactionDB::Open(..., &txn_db_)
| db_ is still nullptr
| db_->GetSnapshot() // segfault
| db_ = txn_db_
V
```
Akanksha Mahajan [Tue, 14 Jun 2022 01:07:24 +0000 (18:07 -0700)]
Implement AbortIO using io_uring (#10125)
Summary:
Implement AbortIO in posix using io_uring to cancel any pending read requests submitted. Its cancelled using io_uring_prep_cancel which sets the IORING_OP_ASYNC_CANCEL flag.
To cancel a request, the sqe must have ->addr set to the user_data of the request it wishes to cancel. If the request is cancelled successfully, the original request is completed with -ECANCELED and the cancel request is completed with a result of 0. If the request was already running, the original may or may not complete in error. The cancel request will complete with -EALREADY for that case. And finally, if the request to cancel wasn't found, the cancel request is completed with -ENOENT.
Mark Callaghan [Mon, 13 Jun 2022 23:24:32 +0000 (16:24 -0700)]
Increase num_levels for universal from 8 to 40 (#10158)
Summary:
See https://github.com/facebook/rocksdb/issues/10082 for more details. Trivial move
isn't done for universal when compaction is from L0 into L0. So a too small value for
num_levels with db_bench means there will be fewer trivial moves with universal and
that means that write-amp will increase.
Peter Dillinger [Mon, 13 Jun 2022 18:08:50 +0000 (11:08 -0700)]
Document design/specification bugs with auto_prefix_mode (#10144)
Summary:
auto_prefix_mode is designed to use prefix filtering in a
particular "safe" set of cases where the upper bound and the seek key
have different prefixes: where the upper bound is the "same length
immediate successor". These conditions are not sufficient to guarantee
the same iteration results as total_order_seek if the DB contains
"short" keys, less than the "full" (maximum) prefix length.
We are not simply disabling the optimization in these successor cases
because it is likely that users are essentially getting what they want
out of existing usage. Especially if users are constructing successor
bounds with the intention of doing a prefix-bounded seek, the existing
behavior is more expected than the total_order_seek behavior.
Consequently, for now we reconcile the bad specification of behavior by
documenting the existing mismatch with total_order_seek.
A closely related issue affects hypothetical comparators like
ReverseBytewiseComparator: if they "correctly" implement
IsSameLengthImmediateSuccessor, auto_prefix_mode could omit more
entries (other than "short" keys noted above). Luckily, the built-in
ReverseBytewiseComparator has an "incorrect" implementation of
IsSameLengthImmediateSuccessor that effectively prevents prefix
optimization and, thus, the bug. This is now documented as a new
constraint on IsSameLengthImmediateSuccessor, and the implementation
tweaked to be simply "safe" rather than "incorrect".
This change also includes unit test updates to demonstrate the above
issues. (Test was cleaned up for readability and simplicity.)
Intended follow-up:
* Tweak documented axioms for prefix_extractor (more details then)
* Consider some sort of fix for this case. I don't know what that would
look like without breaking the performance of existing code. Perhaps
if all keys in an SST file have prefixes that are "full length," we can track
that fact and use it to allow optimization with the "same length
immediate successor", but that would only apply to new files.
* Consider a better system of specifying prefix bounds
Summary:
FastLRUCache now only supports 16B keys. The tests have changed to reflect this.
Because the unit tests were designed for caches that accept any string as keys, some tests are no longer compatible with FastLRUCache. We have disabled those for runs with FastLRUCache. (We could potentially change all tests to use 16B keys, but we don't because the cache public API does not require this.)
Yanqin Jin [Fri, 10 Jun 2022 23:07:03 +0000 (16:07 -0700)]
Snapshots with user-specified timestamps (#9879)
Summary:
In RocksDB, keys are associated with (internal) sequence numbers which denote when the keys are written
to the database. Sequence numbers in different RocksDB instances are unrelated, thus not comparable.
It is nice if we can associate sequence numbers with their corresponding actual timestamps. One thing we can
do is to support user-defined timestamp, which allows the applications to specify the format of custom timestamps
and encode a timestamp with each key. More details can be found at https://github.com/facebook/rocksdb/wiki/User-defined-Timestamp-%28Experimental%29.
This PR provides a different but complementary approach. We can associate rocksdb snapshots (defined in
https://github.com/facebook/rocksdb/blob/7.2.fb/include/rocksdb/snapshot.h#L20) with **user-specified** timestamps.
Since a snapshot is essentially an object representing a sequence number, this PR establishes a bi-directional mapping between sequence numbers and timestamps.
In the past, snapshots are usually taken by readers. The current super-version is grabbed, and a `rocksdb::Snapshot`
object is created with the last published sequence number of the super-version. You can see that the reader actually
has no good idea of what timestamp to assign to this snapshot, because by the time the `GetSnapshot()` is called,
an arbitrarily long period of time may have already elapsed since the last write, which is when the last published
sequence number is written.
This observation motivates the creation of "timestamped" snapshots on the write path. Currently, this functionality is
exposed only to the layer of `TransactionDB`. Application can tell RocksDB to create a snapshot when a transaction
commits, effectively associating the last sequence number with a timestamp. It is also assumed that application will
ensure any two snapshots with timestamps should satisfy the following:
```
snapshot1.seq < snapshot2.seq iff. snapshot1.ts < snapshot2.ts
```
If the application can guarantee that when a reader takes a timestamped snapshot, there is no active writes going on
in the database, then we also allow the user to use a new API `TransactionDB::CreateTimestampedSnapshot()` to create
a snapshot with associated timestamp.
Code example
```cpp
// Create a timestamped snapshot when committing transaction.
txn->SetCommitTimestamp(100);
txn->SetSnapshotOnNextOperation();
txn->Commit();
// A wrapper API for convenience
Status Transaction::CommitAndTryCreateSnapshot(
std::shared_ptr<TransactionNotifier> notifier,
TxnTimestamp ts,
std::shared_ptr<const Snapshot>* ret);
// Create a timestamped snapshot if caller guarantees no concurrent writes
std::pair<Status, std::shared_ptr<const Snapshot>> snapshot = txn_db->CreateTimestampedSnapshot(100);
```
The snapshots created in this way will be managed by RocksDB with ref-counting and potentially shared with
other readers. We provide the following APIs for readers to retrieve a snapshot given a timestamp.
```cpp
// Return the timestamped snapshot correponding to given timestamp. If ts is
// kMaxTxnTimestamp, then we return the latest timestamped snapshot if present.
// Othersise, we return the snapshot whose timestamp is equal to `ts`. If no
// such snapshot exists, then we return null.
std::shared_ptr<const Snapshot> TransactionDB::GetTimestampedSnapshot(TxnTimestamp ts) const;
// Return the latest timestamped snapshot if present.
std::shared_ptr<const Snapshot> TransactionDB::GetLatestTimestampedSnapshot() const;
```
We also provide two additional APIs for stats collection and reporting purposes.
```cpp
Status TransactionDB::GetAllTimestampedSnapshots(
std::vector<std::shared_ptr<const Snapshot>>& snapshots) const;
// Return timestamped snapshots whose timestamps fall in [ts_lb, ts_ub) and store them in `snapshots`.
Status TransactionDB::GetTimestampedSnapshots(
TxnTimestamp ts_lb,
TxnTimestamp ts_ub,
std::vector<std::shared_ptr<const Snapshot>>& snapshots) const;
```
To prevent the number of timestamped snapshots from growing infinitely, we provide the following API to release
timestamped snapshots whose timestamps are older than or equal to a given threshold.
```cpp
void TransactionDB::ReleaseTimestampedSnapshotsOlderThan(TxnTimestamp ts);
```
Before shutdown, RocksDB will release all timestamped snapshots.
Comparison with user-defined timestamp and how they can be combined:
User-defined timestamp persists every key with a timestamp, while timestamped snapshots maintain a volatile
mapping between snapshots (sequence numbers) and timestamps.
Different internal keys with the same user key but different timestamps will be treated as different by compaction,
thus a newer version will not hide older versions (with smaller timestamps) unless they are eligible for garbage collection.
In contrast, taking a timestamped snapshot at a certain sequence number and timestamp prevents all the keys visible in
this snapshot from been dropped by compaction. Here, visible means (seq < snapshot and most recent).
The timestamped snapshot supports the semantics of reading at an exact point in time.
Timestamped snapshots can also be used with user-defined timestamp.
Peter Dillinger [Fri, 10 Jun 2022 15:51:45 +0000 (08:51 -0700)]
Fix bug with kHashSearch and changing prefix_extractor with SetOptions (#10128)
Summary:
When opening an SST file created using index_type=kHashSearch,
the *current* prefix_extractor would be saved, and used with hash index
if the *new current* prefix_extractor at query time is compatible with
the SST file. This is a problem if the prefix_extractor at SST open time
is not compatible but SetOptions later changes (back) to one that is
compatible.
This change fixes that by using the known compatible (or missing) prefix
extractor we save for use with prefix filtering. Detail: I have moved the
InternalKeySliceTransform wrapper to avoid some indirection and remove
unnecessary fields.
Test Plan:
expanded unit test (using some logic from https://github.com/facebook/rocksdb/issues/10122) that fails
before fix and probably covers some other previously uncovered cases.
Yu Zhang [Fri, 10 Jun 2022 15:21:08 +0000 (08:21 -0700)]
Return try again when full_history_ts_low is higher than requested ts (#10126)
Summary:
This PR helps handle the race condition mentioned in this comment thread: https://github.com/facebook/rocksdb/pull/7884#discussion_r572402281 In case where actual full_history_ts_low is higher than the user's requested ts, return a try again message so they don't have the misconception that data between [ts, full_history_ts_low) is kept.
Test Plan:
```
$COMPILE_WITH_ASAN=1 make -j24 all
$./db_with_timestamp_basic_test --gtest_filter=UpdateFullHistoryTsLowTest.ConcurrentUpdate
$ make -j24 check
```
Summary:
As seen in https://github.com/facebook/rocksdb/issues/10137, simply churning the cache key hashes (e.g.
by changing the raw cache keys) could trigger failure in this test, due
to possibility of some cache shard exceeding its portion of capacity
and evicting entries. Updated the test to be less fragile by using
greater margins, and added a pre-check for evictions, which doesn't
manifest as a race condition, before the main check that can race.
Also added stack trace handler to cache_test for debugging.
Test Plan:
test thousands of iterations with gtest-parallel, including
with changes in https://github.com/facebook/rocksdb/issues/10137 that were surfacing the problem. Pre-check
without the fix would always fail with https://github.com/facebook/rocksdb/issues/10137
Bo Wang [Fri, 10 Jun 2022 00:13:13 +0000 (17:13 -0700)]
Update jemalloc version for platform009 (#10143)
Summary:
Update jemalloc version for platform009. Current one is a bit old and the new one can bring some quick wins (e.g. new heap profiling features on devserver).
Fix bug for WalManager with compressed WAL (#10130)
Summary:
RocksDB uses WalManager to manage WAL files. In WalManager::ReadFirstLine(), the assumption is that reading the first record of a valid WAL file will return OK status and set the output sequence to non-zero value.
This assumption has been broken by WAL compression which writes a `kSetCompressionType` record which is not associated with any sequence number.
Consequently, WalManager::GetSortedWalsOfType() will skip these WALs and not return them to caller, e.g. Checkpoint, Backup, causing the operations to fail.
Mark Callaghan [Wed, 8 Jun 2022 16:23:36 +0000 (09:23 -0700)]
Fix parsing of db_bench output (#10124)
Summary:
A recent diff add a few more fields to one of the db_bench output lines that gets parsed.
This diff updates tools/benchmark.sh to handle that.
Yanqin Jin [Wed, 8 Jun 2022 04:07:47 +0000 (21:07 -0700)]
Update test for secondary instance in stress test (#10121)
Summary:
This PR updates secondary instance testing in stress test by default.
A background thread will be started (disabled by default), running a secondary instance tailing the logs of the primary.
Periodically (every 1 sec), this thread calls `TryCatchUpWithPrimary()` and uses point lookup or range scan
to read some random keys with only very basic verification to make sure no assertion failure is triggered.
Thanks to https://github.com/facebook/rocksdb/issues/10061 , we can enable secondary instance when user-defined timestamp is enabled.
Also removed a less useful test configuration, `secondary_catch_up_one_in`. This is very similar to the periodic
catch-up.
In the last commit, I decided not to enable it now, but just update the tests, since secondary instance does not
work well when the underlying file is renamed by primary, e.g. SstFileManager.
Test Plan:
```
TEST_TMPDIR=/dev/shm/rocksdb make crash_test
TEST_TMPDIR=/dev/shm/rocksdb make crash_test_with_ts
TEST_TMPDIR=/dev/shm/rocksdb make crash_test_with_atomic_flush
```
The command was using default values for `ingest_external_file_width`
(1000) and `log2_keys_per_lock` (2). The expected number of locks needed
to update those keys is then (1000 / 2^2) = 250, which is above the 0x40 (64)
limit. This PR reduces the default value of `ingest_external_file_width`
to 100 so the expected number of locks is 25, which is within the limit.
gitbw95 [Tue, 7 Jun 2022 18:57:12 +0000 (11:57 -0700)]
Add unit test to verify that the dynamic priority can be passed from compaction to FS (#10088)
Summary:
**Summary:**
Add unit tests to verify that the dynamic priority can be passed from compaction to FS. Compaction reads&writes and other DB reads&writes share the same read&write paths to FSRandomAccessFile or FSWritableFile, so a MockTestFileSystem is added to replace the default filesystem from Env to intercept and verify the io_priority. To prepare the compaction input files, use the default filesystem from Env. To test the io priority of the compaction reads and writes, db_options_.fs is set as MockTestFileSystem.
zczhu [Tue, 7 Jun 2022 16:49:31 +0000 (09:49 -0700)]
Handle "NotSupported" status by default implementation of Close() in … (#10127)
Summary:
The default implementation of Close() function in Directory/FSDirectory classes returns `NotSupported` status. However, we don't want operations that worked in older versions to begin failing after upgrading when run on FileSystems that have not implemented Directory::Close() yet. So we require the upper level that calls Close() function should properly handle "NotSupported" status instead of treating it as an error status.
Summary:
As pointed out by [https://github.com/facebook/rocksdb/pull/8351#discussion_r645765422](https://github.com/facebook/rocksdb/pull/8351#discussion_r645765422), check `manual_compaction_paused` and `manual_compaction_canceled` can be reduced by setting `*canceled` to be true in `DisableManualCompaction()` and `*canceled` to be false in the last time calling `EnableManualCompaction()`.
Changed Tests: The origin `DBTest2.PausingManualCompaction1` uses a callback function to increase `manual_compaction_paused` and the origin CompactionJob/CompactionIterator with `manual_compaction_paused` can detect this. I changed the callback function so that it sets `*canceled` as true if `canceled` is not `nullptr` (to notify CompactionJob/CompactionIterator the compaction has been canceled).
Test Plan: This change does not introduce new features, but some slight difference in compaction implementation. Run the same manual compaction unit tests as before (e.g., PausingManualCompaction[1-4], CancelManualCompaction[1-2], CancelManualCompactionWithListener in db_test2, and db_compaction_test).
Yu Zhang [Mon, 6 Jun 2022 21:36:22 +0000 (14:36 -0700)]
Return "invalid argument" when read timestamp is too old (#10109)
Summary:
With this change, when a given read timestamp is smaller than the column-family's full_history_ts_low, Get(), MultiGet() and iterators APIs will return Status::InValidArgument().
Test plan
```
$COMPILE_WITH_ASAN=1 make -j24 all
$./db_with_timestamp_basic_test --gtest_filter=DBBasicTestWithTimestamp.UpdateFullHistoryTsLow
$ make -j24 check
```
zczhu [Mon, 6 Jun 2022 21:27:31 +0000 (14:27 -0700)]
Fix default implementaton of close() function for Directory/FSDirecto… (#10123)
Summary:
As pointed by anand1976 in his [comment](https://github.com/facebook/rocksdb/pull/10049#pullrequestreview-994255819), previous implementation (adding Close() function in Directory/FSDirectory class) is not backward-compatible. And we mistakenly added the default implementation `return Status::NotSupported("Close")` or `return IOStatus::NotSupported("Close")` in WritableFile class in this [pull request](https://github.com/facebook/rocksdb/pull/10101). This pull request fixes the above issue.
Fix overflow bug in standard deviation computation. (#10100)
Summary:
There was an overflow bug when computing the variance in the HistogramStat class.
This manifests, for instance, when running cache_bench with default arguments. This executes 32M lookups/inserts/deletes in a block cache, and then computes (among other things) the variance of the latencies. The variance is computed as ``variance = (cur_sum_squares * cur_num - cur_sum * cur_sum) / (cur_num * cur_num)``, where ``cum_sum_squares`` is the sum of the squares of the samples, ``cur_num`` is the number of samples, and ``cur_sum`` is the sum of the samples. Because the median latency in a typical run is around 3800 nanoseconds, both the ``cur_sum_squares * cur_num`` and ``cur_sum * cur_sum`` terms overflow as uint64_t.
Peter Dillinger [Mon, 6 Jun 2022 18:16:12 +0000 (11:16 -0700)]
Refactor: Add BlockTypes to make them imply C++ type in block cache (#10098)
Summary:
We have three related concepts:
* BlockType: an internal enum conceptually indicating a type of SST file
block
* CacheEntryRole: a user-facing enum for categorizing block cache entries,
which is also involved in associated cache entries with an appropriate
deleter. Can include categories for non-block cache entries (e.g. memory
reservations).
* TBlocklike: a C++ type for the actual type behind a void* cache entry.
We had some existing code ugliness because BlockType did not imply
TBlocklike, because of various kinds of "filter" block. This refactoring
fixes that with new BlockTypes.
Alan Paxton [Sat, 4 Jun 2022 16:31:47 +0000 (09:31 -0700)]
CI Benchmarking with CircleCI Runner and OpenSearch Dashboard (EB 1088) (#9723)
Summary:
CircleCI runner based benchmarking. A runner is a dedicate machine configured for CircleCI to perform work on. Our work is a repeatable benchmark, the `benchmark-linux` job in `config.yml`
A runner, in CircleCI terminology, is a machine that is managed by the client (us) rather than running on CircleCI resources in the cloud. This means that we define and configure the iron, and that therefore the performance is repeatable and predictable. Which is what we need for performance regression benchmarking.
On a time schedule (or on commit, during branch development) benchmarks are set off on the runner, and then a script is run `benchmark_log_tool.py` which parses the benchmark output and pushes it into a pre-configured OpenSearch document connected to an OpenSearch dashboard. Members of the team can examine benchmark performance changes on the dashboard.
As time progresses we can add different benchmarks to the suite which gets run.
Levi Tamasi [Sat, 4 Jun 2022 03:54:48 +0000 (20:54 -0700)]
Add wide column serialization primitives (#9915)
Summary:
The patch adds some low-level logic that can be used to serialize/deserialize
a sorted vector of wide columns to/from a simple binary searchable string
representation. Currently, there is no user-facing API; this will be implemented in
subsequent stages.
Yanqin Jin [Sat, 4 Jun 2022 03:00:42 +0000 (20:00 -0700)]
Point-lookup returns timestamps of Delete and SingleDelete (#10056)
Summary:
If caller specifies a non-null `timestamp` argument in `DB::Get()` or a non-null `timestamps` in `DB::MultiGet()`,
RocksDB will return the timestamps of the point tombstones.
Hui Xiao [Sat, 4 Jun 2022 02:42:22 +0000 (19:42 -0700)]
Increase ChargeTableReaderTest/ChargeTableReaderTest.Basic error tolerance rate from 1% to 5% (#10113)
Summary:
**Context:**
https://github.com/facebook/rocksdb/pull/9748 added support to charge table reader memory to block cache. In the test `ChargeTableReaderTest/ChargeTableReaderTest.Basic`, it estimated the table reader memory, calculated the expected number of table reader opened based on this estimation and asserted this number with actual number. The expected number of table reader opened calculated based on estimated table reader memory will not be 100% accurate and should have tolerance for error. It was previously set to 1% and recently encountered an assertion failure that `(opened_table_reader_num) <= (max_table_reader_num_capped_upper_bound), actual: 375 or 376 vs 374` where `opened_table_reader_num` is the actual opened one and `max_table_reader_num_capped_upper_bound` is the estimated opened one (=371 * 1.01). I believe it's safe to increase error tolerance from 1% to 5% hence there is this PR.
Levi Tamasi [Fri, 3 Jun 2022 23:35:13 +0000 (16:35 -0700)]
Fix some bugs in verify_random_db.sh (#10112)
Summary:
The patch attempts to fix three bugs in `verify_random_db.sh`:
1) https://github.com/facebook/rocksdb/pull/9937 changed the default for
`--try_load_options` to true in the script's use case, so we have to
explicitly set it to false if the corresponding argument of the script
is 0. This should fix the issue we've been seeing with our forward
compatibility tests where 7.3 is unable to open a database created by
the version on main after adding a new configuration option.
2) The script seems to support two "extra parameters"; however,
in practice, if the second one was set, only that one was passed on to
`ldb`. Now both get forwarded.
3) When running the `diff` command, the base DB directory was passed as
the second argument instead of the file containing the `ldb` output
(this actually seems to work, probably accidentally though).
Test Plan:
- Temporarily add an ``assert(false)`` in the execution path that sets up the FastLRUCache. Run ``make -j24 cache_bench``. Then test the appropriate code is used by running ``./cache_bench -cache_type=fast_lru_cache`` and checking that the assert is called. Repeat for LRUCache.
- Verify that FastLRUCache (currently a clone of LRUCache) has similar latency distribution than LRUCache, by comparing the outputs of ``./cache_bench -cache_type=fast_lru_cache`` and ``./cache_bench -cache_type=lru_cache``.
zczhu [Fri, 3 Jun 2022 19:53:28 +0000 (12:53 -0700)]
Add default impl to dir close (#10101)
Summary:
As pointed by anand1976 in his [comment](https://github.com/facebook/rocksdb/pull/10049#pullrequestreview-994255819), previous implementation is not backward-compatible. In this implementation, the default implementation `return Status::NotSupported("Close")` or `return IOStatus::NotSupported("Close")` is added for `Close()` function for `*Directory` classes.
Test Plan:
- Temporarily add an ``assert(false)`` in the execution path that sets up the FastLRUCache. Run ``make -j24 db_bench``. Then test the appropriate code is used by running ``./db_bench -cache_type=fast_lru_cache`` and checking that the assert is called. Repeat for LRUCache.
- Verify that FastLRUCache (currently a clone of LRUCache) produces similar benchmark data than LRUCache, by comparing the outputs of ``./db_bench -benchmarks=fillseq,fillrandom,readseq,readrandom -cache_type=fast_lru_cache`` and ``./db_bench -benchmarks=fillseq,fillrandom,readseq,readrandom -cache_type=lru_cache``.
Yanqin Jin [Fri, 3 Jun 2022 17:22:52 +0000 (10:22 -0700)]
Temporarily disable wal compression (#10108)
Summary:
Will re-enable after fixing the bug in https://github.com/facebook/rocksdb/issues/10099 and https://github.com/facebook/rocksdb/issues/10097.
Right now, the priority is https://github.com/facebook/rocksdb/issues/10087, but the bug in WAL compression prevents the mini crash test from passing.
Mark Callaghan [Fri, 3 Jun 2022 15:20:10 +0000 (08:20 -0700)]
Enhance to support more tuning options, and universal and integrated… (#9704)
Summary:
… BlobDB for all tests
This does two big things:
* provides more tuning options
* supports universal and integrated BlobDB for all of the benchmarks that are leveled-only
It does several smaller things, and I will list a few
* sets l0_slowdown_writes_trigger which wasn't set before this diff.
* improves readability in report.tsv by using smaller field names in the header
* adds more columns to report.tsv
zczhu [Fri, 3 Jun 2022 04:52:35 +0000 (21:52 -0700)]
Add comments/permit unchecked error to close_db_dir pull requests (#10093)
Summary:
In [close_db_dir](https://github.com/facebook/rocksdb/pull/10049) pull request, some merging conflicts occurred (some comments and one line `s.PermitUncheckedError()` are missing). This pull request aims to put them back.
Gang Liao [Fri, 3 Jun 2022 03:04:33 +0000 (20:04 -0700)]
Make it possible to enable blob files starting from a certain LSM tree level (#10077)
Summary:
Currently, if blob files are enabled (i.e. `enable_blob_files` is true), large values are extracted both during flush/recovery (when SST files are written into level 0 of the LSM tree) and during compaction into any LSM tree level. For certain use cases that have a mix of short-lived and long-lived values, it might make sense to support extracting large values only during compactions whose output level is greater than or equal to a specified LSM tree level (e.g. compactions into L1/L2/... or above). This could reduce the space amplification caused by large values that are turned into garbage shortly after being written at the price of some write amplification incurred by long-lived values whose extraction to blob files is delayed.
In order to achieve this, we would like to do the following:
- Add a new configuration option `blob_file_starting_level` (default: 0) to `AdvancedColumnFamilyOptions` (and `MutableCFOptions` and extend the related logic)
- Instantiate `BlobFileBuilder` in `BuildTable` (used during flush and recovery, where the LSM tree level is L0) and `CompactionJob` iff `enable_blob_files` is set and the LSM tree level is `>= blob_file_starting_level`
- Add unit tests for the new functionality, and add the new option to our stress tests (`db_stress` and `db_crashtest.py` )
- Add the new option to our benchmarking tool `db_bench` and the BlobDB benchmark script `run_blob_bench.sh`
- Add the new option to the `ldb` tool (see https://github.com/facebook/rocksdb/wiki/Administration-and-Data-Access-Tool)
- Ideally extend the C and Java bindings with the new option
- Update the BlobDB wiki to document the new option.
Gang Liao [Thu, 2 Jun 2022 02:40:26 +0000 (19:40 -0700)]
Support specifying blob garbage collection parameters when CompactRange() (#10073)
Summary:
Garbage collection is generally controlled by the BlobDB configuration options `enable_blob_garbage_collection` and `blob_garbage_collection_age_cutoff`. However, there might be use cases where we would want to temporarily override these options while performing a manual compaction. (One use case would be doing a full key-space manual compaction with full=100% garbage collection age cutoff in order to minimize the space occupied by the database.) Our goal here is to make it possible to override the configured GC parameters when using the `CompactRange` API to perform manual compactions. This PR would involve:
- Extending the `CompactRangeOptions` structure so clients can both force-enable and force-disable GC, as well as use a different cutoff than what's currently configured
- Storing whether blob GC should actually be enabled during a certain manual compaction and the cutoff to use in the `Compaction` object (considering the above overrides) and passing it to `CompactionIterator` via `CompactionProxy`
- Updating the BlobDB wiki to document the new options.
Zichen Zhu [Thu, 2 Jun 2022 01:03:34 +0000 (18:03 -0700)]
Explicitly closing all directory file descriptors (#10049)
Summary:
Currently, the DB directory file descriptor is left open until the deconstruction process (`DB::Close()` does not close the file descriptor). To verify this, comment out the lines between `db_ = nullptr` and `db_->Close()` (line 512, 513, 514, 515 in ldb_cmd.cc) to leak the ``db_'' object, build `ldb` tool and run
```
strace --trace=open,openat,close ./ldb --db=$TEST_TMPDIR --ignore_unknown_options put K1 V1 --create_if_missing
```
There is one directory file descriptor that is not closed in the strace log.
Test Plan: Add a new unit test DBBasicTest.DBCloseAllDirectoryFDs: Open a database with different WAL directory and three different data directories, and all directory file descriptors should be closed after calling Close(). Explicitly call Close() after a directory file descriptor is not used so that the counter of directory open and close should be equivalent.
Add support for FastLRUCache in stress and crash tests. (#10081)
Summary:
Stress tests can run with the experimental FastLRUCache. Crash tests randomly choose between LRUCache and FastLRUCache.
Since only LRUCache supports a secondary cache, we validate the `--secondary_cache_uri` and `--cache_type` flags---when `--secondary_cache_uri` is set, the `--cache_type` is set to `lru_cache`.
Test Plan:
- To test that the FastLRUCache is used and the stress test runs successfully, run `make -j24 CRASH_TEST_EXT_ARGS=—duration=960 blackbox_crash_test_with_atomic_flush`. The cache type should sometimes be `fast_lru_cache`.
- To test the flag validation, run `make -j24 CRASH_TEST_EXT_ARGS="--duration=960 --secondary_cache_uri=x" blackbox_crash_test_with_atomic_flush` multiple times. The test will always be aborted (which is okay). Check that the cache type is always `lru_cache`.
Jay Zhuang [Wed, 1 Jun 2022 22:33:22 +0000 (15:33 -0700)]
Get current LogFileNumberSize the same as log_writer (#10086)
Summary:
`db_impl.alive_log_files_` is used to track the WAL size in `db_impl.logs_`.
Get the `LogFileNumberSize` obj in `alive_log_files_` the same time as `log_writer` to keep them consistent.
For this issue, it's not safe to do `deque::reverse_iterator::operator*` and `deque::pop_front()` concurrently,
so remove the tail cache.
Peter Dillinger [Wed, 1 Jun 2022 18:02:27 +0000 (11:02 -0700)]
Reduce risk of backup or checkpoint missing a WAL file (#10083)
Summary:
We recently saw a case in crash test in which a WAL file in the
middle of the list of live WALs was not included in the backup, so the
DB was not openable due to missing WAL. We are not sure why, but this
change should at least turn that into a backup-time failure by ensuring
all the WAL files expected by the manifest (according to VersionSet) are
included in `GetSortedWalFiles()` (used by `GetLiveFilesStorageInfo()`,
`BackupEngine`, and `Checkpoint`)
Related: to maximize the effectiveness of
track_and_verify_wals_in_manifest with GetSortedWalFiles() during
checkpoint/backup, we will now sync WAL in GetLiveFilesStorageInfo()
when track_and_verify_wals_in_manifest=true.
Persist the new MANIFEST after successfully syncing the new WAL during recovery (#9922)
Summary:
In case of non-TransactionDB and avoid_flush_during_recovery = true, RocksDB won't
flush the data from WAL to L0 for all column families if possible. As a
result, not all column families can increase their log_numbers, and
min_log_number_to_keep won't change.
For transaction DB (.allow_2pc), even with the flush, there may be old WAL files that it must not delete because they can contain data of uncommitted transactions and min_log_number_to_keep won't change.
If we persist a new MANIFEST with
advanced log_numbers for some column families, then during a second
crash after persisting the MANIFEST, RocksDB will see some column
families' log_numbers larger than the corrupted wal, and the "column family inconsistency" error will be hit, causing recovery to fail.
As a solution, RocksDB will persist the new MANIFEST after successfully syncing the new WAL.
If a future recovery starts from the new MANIFEST, then it means the new WAL is successfully synced. Due to the sentinel empty write batch at the beginning, kPointInTimeRecovery of WAL is guaranteed to go after this point.
If future recovery starts from the old MANIFEST, it means the writing the new MANIFEST failed. We won't have the "SST ahead of WAL" error.
Currently, RocksDB DB::Open() may creates and writes to two new MANIFEST files even before recovery succeeds. This PR buffers the edits in a structure and writes to a new MANIFEST after recovery is successful
Yanqin Jin [Tue, 31 May 2022 16:36:32 +0000 (09:36 -0700)]
Deflake unit test BackupEngineTest.Concurrency (#10069)
Summary:
After https://github.com/facebook/rocksdb/issues/9984, BackupEngineTest.Concurrency becomes flaky.
During DB::Open(), someone else can rename/remove the LOG file, causing
this thread's `CreateLoggerFromOptions()` to fail. The reason is that the operation sequence
of "FileExists -> Rename" is not atomic. It's possible that a FileExists() returns OK, but the file
gets deleted before Rename(), causing the latter to return IOError with PathNotFound subcode.
Although it's not encouraged to concurrently modify the contents of the directories managed by
the database instance in this case, we can still perform some simple handling to make DB::Open()
more robust. In this case, we can check if a racing thread has deleted the original LOG file, we can
allow this thread to continue creating a new LOG file.
Summary:
Fix the unittest `ExternalSSTFileBasicTest.StableSnapshotWhileLoggingToManifest` introduced in https://github.com/facebook/rocksdb/issues/10051 that is failing.
Andrea Pappacoda [Mon, 30 May 2022 19:46:40 +0000 (12:46 -0700)]
build: fix pkg-config file generation (#9953)
Summary:
- Instead of hardcoding "lib" and "include" in `libdir` and `includedir`, use the values from [`GNUInstallDirs`](https://cmake.org/cmake/help/latest/module/GNUInstallDirs.html).
- Use `PROJECT_DESCRIPTION` and `PROJECT_HOMEPAGE_URL` instead of their
`CMAKE_` conterparts to fix pkg-config generation when rocksdb is not the top-level project (see [`project()`](https://cmake.org/cmake/help/latest/command/project.html)).
- Drop explicit `CMAKE_CURRENT_SOURCE_DIR` and `CMAKE_CURRENT_BINARY_DIR` in [`configure_file()`](https://cmake.org/cmake/help/latest/command/configure_file.html) as that's implied by default (and quite intuitive).
See https://github.com/facebook/rocksdb/issues/9945 CC: trynity
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9953
projects / rocksdb.git / log