Summary:
Based on my recent findings (posted in our internal group), if we use fallocate without KEEP_SIZE flag, we get superior performance of fdatasync() in append-only workloads.
This diff provides an option for user to not use KEEP_SIZE flag, thus optimizing his sync performance by up to 2x-3x.
At one point we also just called posix_fallocate instead of fallocate, which isn't very fast: http://code.woboq.org/userspace/glibc/sysdeps/posix/posix_fallocate.c.html (tl;dr it manually writes out zero bytes to allocate storage). This diff also fixes that, by first calling fallocate and then posix_fallocate if fallocate is not supported.
Test Plan: make check
Reviewers: dhruba, sdong, haobo, ljin
Reviewed By: dhruba
CC: leveldb
Differential Revision: https://reviews.facebook.net/D16761
Summary:
Original Summary:
Yesterday, @ljin and I were debugging various db_stress issues. We suspected one of them happens when we concurrently call NewIterator without prefix_seek on HashSkipList. This test demonstrates it.
Update:
Arena is not thread-safe!! When creating a new full iterator, we *have* to create a new arena, otherwise we're doomed.
Test Plan: SIGSEGV and assertion-throwing test now works!
Reviewers: ljin, haobo, sdong
Reviewed By: sdong
CC: leveldb, ljin
Differential Revision: https://reviews.facebook.net/D16857
Summary:
This is is based on https://reviews.facebook.net/D15027. It's not finished but I would like to give a prototype to avoid arena over-allocation while making better use of the already allocated memory blocks.
Instead of check approximate memtable size, we will take a deeper look at the arena, which incorporate essential idea that @sdong suggests: flush when arena has allocated its last and the last is "almost full"
Test Plan: N/A
Reviewers: haobo, sdong
Reviewed By: sdong
CC: leveldb, sdong
Differential Revision: https://reviews.facebook.net/D15051
Summary:
@igor pointed out that there is a potential data race because of the way we use the newly introduced LogBuffer. After "bg_compaction_scheduled_--" or "bg_flush_scheduled_--", they can both become 0. As soon as the lock is released after that, DBImpl's deconstructor can go ahead and deconstruct all the states inside DB, including the info_log object hold in a shared pointer of the options object it keeps. At that point it is not safe anymore to continue using the info logger to write the delayed logs.
With the patch, lock is released temporarily for log buffer to be flushed before "bg_compaction_scheduled_--" or "bg_flush_scheduled_--". In order to make sure we don't miss any pending flush or compaction, a new flag bg_schedule_needed_ is added, which is set to be true if there is a pending flush or compaction but not scheduled because of the max thread limit. If the flag is set to be true, the scheduling function will be called before compaction or flush thread finishes.
Thanks @igor for this finding!
Test Plan: make all check
Reviewers: haobo, igor
Reviewed By: haobo
CC: dhruba, ljin, yhchiang, igor, leveldb
Differential Revision: https://reviews.facebook.net/D16767
Summary: Add a function to Env so that users can query the waiting queue length of each thread pool
Test Plan: add a test in env_test
Reviewers: haobo
Reviewed By: haobo
CC: dhruba, igor, yhchiang, ljin, nkg-, leveldb
Differential Revision: https://reviews.facebook.net/D16755
Summary:
(1) Fix SanitizeOptions() to also check HashLinkList. The current
dynamic case just happens to work because the 2 classes have the same
layout.
(2) Do not delete SliceTransform object in HashSkipListFactory and
HashLinkListFactory destructor. Reason: SanitizeOptions() enforces
prefix_extractor and SliceTransform to be the same object when
Hash**Factory is used. This makes the behavior strange: when
Hash**Factory is used, prefix_extractor will be released by RocksDB. If
other memtable factory is used, prefix_extractor should be released by
user.
Test Plan: db_bench && make asan_check
Reviewers: haobo, igor, sdong
Reviewed By: igor
CC: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D16587
Summary: Moved LogBuffer class to an internal header. Removed some unneccesary indirection. Enabled log buffer for BackgroundCallFlush. Forced log buffer flush right after Unlock to improve time ordering of info log.
Test Plan: make check; db_bench compare LOG output
Reviewers: sdong
Reviewed By: sdong
CC: leveldb, igor
Differential Revision: https://reviews.facebook.net/D16707
Summary:
If verify_checksums_in_compaction is true, compaction will verify checksums. This is default.
If it's false, compaction doesn't verify checksums. This is useful for in-memory workloads.
Test Plan: corruption_test
Reviewers: dhruba
Reviewed By: dhruba
CC: leveldb
Differential Revision: https://reviews.facebook.net/D16695
Summary:
Add a check at the end of GetImpl to release SuperVersion if it becomes
obsolete. Also do Scrape() inside InstallSuperVersion so it happens more
frequent.
Test Plan:
make all check
running asan_check now
Reviewers: igor, haobo, sdong, dhruba
Reviewed By: haobo
CC: leveldb
Differential Revision: https://reviews.facebook.net/D16641
Summary:
Change to store the return value from ftruncate().
The reason is that ftruncate() has "warn_unused_result" attribute in some environment.
Signed-off-by: Yumikiyo Osanai <yumios.art@gmail.com>
Summary: valgrind reports issues. This patch seems to fix it.
Test Plan: run the tests that fails in valgrind
Reviewers: igor, haobo, kailiu
Reviewed By: kailiu
CC: dhruba, ljin, yhchiang, leveldb
Differential Revision: https://reviews.facebook.net/D16653
Summary:
Blocks allocated with fallocate will take extra space on disk even if they are unused and the file is close.
Now we remove the extra blocks at the end of the file by calling `ftruncate`.
Test Plan: added a test to env_test
Reviewers: dhruba
Reviewed By: dhruba
CC: leveldb
Differential Revision: https://reviews.facebook.net/D16647
Summary:
With the use of tmpfs or ramfs, unit tests related to GetUniqueID()
failed because of the failure from ioctl, which doesn't work with these
fancy file systems at all.
I fixed this issue and make sure all related tests run on the "regular"
storage (disk or flash).
Test Plan: TEST_TMPDIR=/dev/shm make check -j32
Reviewers: igor, dhruba
CC: leveldb
Differential Revision: https://reviews.facebook.net/D16593
Summary: Now while the background thread is picking compactions, it writes out multiple info_logs, especially for universal compaction, which introduces a chance of waiting log writing in mutex, which is bad. To remove this risk, write all those info logs to a buffer and flush it after releasing the mutex.
Test Plan:
make all check
check the log lines while running some tests that trigger compactions.
Reviewers: haobo, igor, dhruba
Reviewed By: dhruba
CC: i.am.jin.lei, dhruba, yhchiang, leveldb, nkg-
Differential Revision: https://reviews.facebook.net/D16515
Summary:
Currently, there is no easy way for user to change log level of info log. Add a parameter in options to specify that.
Also make the default level to INFO level. Removing the [INFO] tag if it is INFO level as I don't want to cause performance regression. (add [LOG] means another mem-copy and string formatting).
Test Plan:
make all check
manual check the levels work as expected.
Reviewers: dhruba, yhchiang
Reviewed By: yhchiang
CC: dhruba, igor, i.am.jin.lei, ljin, haobo, leveldb
Differential Revision: https://reviews.facebook.net/D16563
Summary:
Add helper function to print perf context data in db_bench if enabled.
I didn't find any code that actually exports perf context data. Not sure
if I missed anything
Test Plan: ran db_bench
Reviewers: haobo, sdong, igor
Reviewed By: igor
CC: leveldb
Differential Revision: https://reviews.facebook.net/D16575
Summary:
this is the key component extracted from diff: https://reviews.facebook.net/D14271
I separate it to a dedicated patch to make the review easier.
Test Plan: added a unit test and passed it.
Reviewers: haobo, sdong, dhruba
CC: leveldb
Differential Revision: https://reviews.facebook.net/D16245
Summary:
This diff does two things:
(1) Log::Reader does not report a corruption when the last record in a log or manifest file is truncated (meaning that log writer died in the middle of the write). Inherited the code from LevelDB: https://code.google.com/p/leveldb/source/detail?r=269fc6ca9416129248db5ca57050cd5d39d177c8#
(2) Turn off mmap writes for all writes to log and manifest files
(2) is necessary because if we use mmap writes, the last record is not truncated, but is actually filled with zeros, making checksum fail. It is hard to recover from checksum failing.
Test Plan:
Added unit tests from LevelDB
Actually recovered a "corrupted" MANIFEST file.
Reviewers: dhruba, haobo
Reviewed By: haobo
CC: leveldb
Differential Revision: https://reviews.facebook.net/D16119
Summary:
InfoLogLevel test now checks the number of lines of the output log file
instead of the number of bytes in the log file.
This diff fixes the issue that the previous InfoLogLevel test in
auto_roll_logger_test passed in make check but fails when valgrind
is used.
Test Plan: run with make check and valgrind.
Reviewers: kailiu
Reviewed By: kailiu
CC: leveldb
Differential Revision: https://reviews.facebook.net/D16407
Summary: as title
Test Plan:
asan_check
will post results later
Reviewers: haobo, igor, dhruba, sdong
Reviewed By: haobo
CC: leveldb
Differential Revision: https://reviews.facebook.net/D16257
Summary: fix the memory leak that was captured by jenkin build.
Test Plan: ran the valgrind test locally
Reviewers: yhchiang
CC: leveldb
Differential Revision: https://reviews.facebook.net/D16389
Summary:
* Now each Log related function has a variant that takes an additional
argument indicating its log level, which is one of the following:
- DEBUG, INFO, WARN, ERROR, FATAL.
* To ensure backward-compatibility, old version Log functions are kept
unchanged.
* Logger now has a member variable indicating its log level. Any incoming
Log request which log level is lower than Logger's log level will not
be output.
* The output of the newer version Log will be prefixed by its log level.
Test Plan:
Add a LogType test in auto_roll_logger_test.cc
= Sample log output =
2014/02/11-00:03:07.683895 7feded179840 [DEBUG] this is the message to be written to the log file!!
2014/02/11-00:03:07.683898 7feded179840 [INFO] this is the message to be written to the log file!!
2014/02/11-00:03:07.683900 7feded179840 [WARN] this is the message to be written to the log file!!
2014/02/11-00:03:07.683903 7feded179840 [ERROR] this is the message to be written to the log file!!
2014/02/11-00:03:07.683906 7feded179840 [FATAL] this is the message to be written to the log file!!
Reviewers: dhruba, xjin, kailiu
Reviewed By: kailiu
CC: leveldb
Differential Revision: https://reviews.facebook.net/D16071
Summary:
This is not a generic thread local implementation in the sense that it
only takes pointer. But it does support multiple instances per thread
and lets user plugin function to perform cleanup when thread exits or an
instance gets destroyed.
Test Plan: unit test for now
Reviewers: haobo, igor, sdong, dhruba
Reviewed By: igor
CC: leveldb, kailiu
Differential Revision: https://reviews.facebook.net/D16131
Summary: A simple benchmark that simulates WAL append. It can be used to test different platform/file system's performance on WAL.
Test Plan: run it.
Reviewers: haobo, kailiu
Reviewed By: haobo
CC: igor, dhruba, i.am.jin.lei, yhchiang, leveldb, nkg-
Differential Revision: https://reviews.facebook.net/D16239
Summary: Clean up IOErrors so that it only indicates errors talking to device.
Test Plan: make all check
Reviewers: igor, haobo, dhruba, emayanke
Reviewed By: igor
CC: leveldb
Differential Revision: https://reviews.facebook.net/D15831
Summary:
This patch optimized Get() code paths by avoiding malloc of iterators. Iterator creation is moved to mem table rep implementations, where a callback is called when any key is found. This is the same practice as what we do in (SST) table readers.
db_bench result for readrandom following a writeseq, with no compression, single thread and tmpfs, we see throughput improved to 144958 from 139027, about 3%.
Test Plan: make all check
Reviewers: dhruba, haobo, igor
Reviewed By: haobo
CC: leveldb, yhchiang
Differential Revision: https://reviews.facebook.net/D14685
Summary: Nothing major, just an extra return line and posibility of leaking fb in NewRandomRWFile
Test Plan: make check
Reviewers: kailiu, dhruba
Reviewed By: kailiu
CC: leveldb
Differential Revision: https://reviews.facebook.net/D15993
Summary: RocksDB doesn't compile on 32-bit architecture apparently. This is attempt to fix some of 32-bit errors. They are reported here: https://gist.github.com/paxos/8789697
Test Plan: RocksDB still compiles on 64-bit :)
Reviewers: kailiu
Reviewed By: kailiu
CC: leveldb
Differential Revision: https://reviews.facebook.net/D15825
Summary: In PlainTable, use one single byte to represent 8 bytes of internal bytes, if seqID = 0 and it is value type (which should be common for bottom most files). It is to save 7 bytes for uncompressed cases.
Test Plan: make all check
Reviewers: haobo, dhruba, kailiu
Reviewed By: haobo
CC: igor, leveldb
Differential Revision: https://reviews.facebook.net/D15489
Summary:
Easy thing goes first. This patch moves arena to internal dir; based
on which, the coming patch will deal with memtable_rep.
Test Plan: make check
Reviewers: haobo, sdong, dhruba
CC: leveldb
Differential Revision: https://reviews.facebook.net/D15615
Summary:
@dhruba, I'm not sure where we need to sync the directory. I implemented the function in Env() and added the dir sync just after we close the newly created file in the builder.
Should I also add FsyncDir() to new files that get created by a compaction?
Test Plan: Confirmed that FsyncDir is returning Status::OK()
Reviewers: dhruba, haobo
Reviewed By: dhruba
CC: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D14751
Summary: Converting from length prefixed buffer back to internal key costs some CPU but it is not necessary. In this patch, internal keys are pass though the functions so that we don't need to convert back to it.
Test Plan: make all check
Reviewers: haobo, kailiu
Reviewed By: kailiu
CC: igor, leveldb
Differential Revision: https://reviews.facebook.net/D15393
Summary: By removing some includes form options.h and reply on forward declaration, we can more easily reason the dependencies.
Test Plan: make all check
Reviewers: kailiu, haobo, igor, dhruba
Reviewed By: kailiu
CC: leveldb
Differential Revision: https://reviews.facebook.net/D15411
Summary: On a shutdown, freeing memory takes a long time. If we're shutting down, we don't really care about memory leaks. I added a call to Cache that will avoid freeing all objects in cache.
Test Plan:
I created a script to test the speedup and demonstrate how to use the call: https://phabricator.fb.com/P3864368
Clean shutdown took 7.2 seconds, while fast and dirty one took 6.3 seconds. Unfortunately, the speedup is not that big, but should be bigger with bigger block_cache. I have set up the capacity to 80GB, but the script filled up only ~7GB.
Reviewers: dhruba, haobo, MarkCallaghan, xjin
Reviewed By: dhruba
CC: leveldb
Differential Revision: https://reviews.facebook.net/D15069
Summary:
A lot of our code implicitly assumes number_levels to be static. ReduceNumberOfLevels() breaks that assumption. For example, after calling ReduceNumberOfLevels(), DBImpl::NumberLevels() will be different from VersionSet::NumberLevels(). This is dangerous. Thankfully, it's not in public headers and is only used from LDB cmd tool. LDB tool is only using it statically, i.e. it never calls it with running DB instance. With this diff, we make it explicitly static. This way, we can assume number_levels to be immutable and not break assumption that lot of our code is relying upon. LDB tool can still use the method.
Also, I removed the method from a separate file since it breaks filename completition. version_se<TAB> now completes to "version_set." instead of "version_set" (without the dot). I don't see a big reason that the function should be in a different file.
Test Plan: reduce_levels_test
Reviewers: dhruba, haobo, kailiu, sdong
Reviewed By: kailiu
CC: leveldb
Differential Revision: https://reviews.facebook.net/D15303
Summary: We have 3 versions of GetLengthPrefixedSlice() and one of them is no longer in use.
Test Plan: make
Reviewers: sdong, igor, haobo, dhruba
CC: leveldb
Differential Revision: https://reviews.facebook.net/D15399
Summary:
For some reason, D15099 caused a big performance regression: https://fburl.com/16059000
After digging a bit, I figured out that the reason was that std::atomic_uint_fast64_t was allocated in an array. When I switched from an array to vector, the QPS returned to the previous level. I'm not sure why this is happening, but this diff seems to fix the performance regression.
Test Plan: I ran the regression script, observed the performance going back to normal
Reviewers: tnovak, kailiu, haobo
Reviewed By: kailiu
CC: leveldb
Differential Revision: https://reviews.facebook.net/D15375
Summary:
This diff takes an even more aggressive way to inline the functions. A decent rule that I followed is "not inline a function if it is more than 10 lines long."
Normally optimizing code by inline is ugly and hard to control, but since one of our usecase has significant amount of CPU used in functions from coding.cc, I'd like to try this diff out.
Test Plan:
1. the size for some .o file increased a little bit, but most less than 1%. So I think the negative impact of inline is negligible.
2. As the regression test shows (ran for 10 times and I calculated the average number)
Metrics Befor After
========================================================================
rocksdb.build.fillseq.qps 426595 444515 (+4.6%)
rocksdb.build.memtablefillrandom.qps 121739 123110
rocksdb.build.memtablereadrandom.qps 1285103 1280520
rocksdb.build.overwrite.qps 125816 135570 (+9%)
rocksdb.build.readrandom_fillunique_random.qps 285995 296863
rocksdb.build.readrandom_memtable_sst.qps 1027132 1027279
rocksdb.build.readrandom.qps 1041427 1054665
rocksdb.build.readrandom_smallblockcache.qps 1028631 1038433
rocksdb.build.readwhilewriting.qps 918352 914629
Reviewers: haobo, sdong, igor
CC: leveldb
Differential Revision: https://reviews.facebook.net/D15291
Igor Canadi