Summary:
There is a new option called hybrid_mode which, when switched on,
causes HBase style compactions. Files from L0 are
compacted back into L0. This meat of this compaction algorithm
is in PickCompactionHybrid().
All files reside in L0. That means all files have overlapping
keys. Each file has a time-bound, i.e. each file contains a
range of keys that were inserted around the same time. The
start-seqno and the end-seqno refers to the timeframe when
these keys were inserted. Files that have contiguous seqno
are compacted together into a larger file. All files are
ordered from most recent to the oldest.
The current compaction algorithm starts to look for
candidate files starting from the most recent file. It continues to
add more files to the same compaction run as long as the
sum of the files chosen till now is smaller than the next
candidate file size. This logic needs to be debated
and validated.
The above logic should reduce write amplification to a
large extent... will publish numbers shortly.
Test Plan: dbstress runs for 6 hours with no data corruption (tested so far).
Differential Revision: https://reviews.facebook.net/D11289
Summary:
There is a new option called hybrid_mode which, when switched on,
causes HBase style compactions. Files from L0 are
compacted back into L0. This meat of this compaction algorithm
is in PickCompactionHybrid().
All files reside in L0. That means all files have overlapping
keys. Each file has a time-bound, i.e. each file contains a
range of keys that were inserted around the same time. The
start-seqno and the end-seqno refers to the timeframe when
these keys were inserted. Files that have contiguous seqno
are compacted together into a larger file. All files are
ordered from most recent to the oldest.
The current compaction algorithm starts to look for
candidate files starting from the most recent file. It continues to
add more files to the same compaction run as long as the
sum of the files chosen till now is smaller than the next
candidate file size. This logic needs to be debated
and validated.
The above logic should reduce write amplification to a
large extent... will publish numbers shortly.
Test Plan: dbstress runs for 6 hours with no data corruption (tested so far).
Differential Revision: https://reviews.facebook.net/D11289
Summary:
This diff simplifies EnvOptions by treating it as POD, similar to Options.
- virtual functions are removed and member fields are accessed directly.
- StorageOptions is removed.
- Options.allow_readahead and Options.allow_readahead_compactions are deprecated.
- Unused global variables are removed: useOsBuffer, useFsReadAhead, useMmapRead, useMmapWrite
Test Plan: make check; db_stress
Reviewers: dhruba
CC: leveldb
Differential Revision: https://reviews.facebook.net/D11175
Summary: Added the 'score' column to the compaction stats output, which shows the level total size devided by level target size. Could be useful when monitoring compaction decisions...
Test Plan: make check; db_bench
Reviewers: dhruba
CC: leveldb, MarkCallaghan
Differential Revision: https://reviews.facebook.net/D11025
Summary:
This diff introduces a new Merge operation into rocksdb.
The purpose of this review is mostly getting feedback from the team (everyone please) on the design.
Please focus on the four files under include/leveldb/, as they spell the client visible interface change.
include/leveldb/db.h
include/leveldb/merge_operator.h
include/leveldb/options.h
include/leveldb/write_batch.h
Please go over local/my_test.cc carefully, as it is a concerete use case.
Please also review the impelmentation files to see if the straw man implementation makes sense.
Note that, the diff does pass all make check and truly supports forward iterator over db and a version
of Get that's based on iterator.
Future work:
- Integration with compaction
- A raw Get implementation
I am working on a wiki that explains the design and implementation choices, but coding comes
just naturally and I think it might be a good idea to share the code earlier. The code is
heavily commented.
Test Plan: run all local tests
Reviewers: dhruba, heyongqiang
Reviewed By: dhruba
CC: leveldb, zshao, sheki, emayanke, MarkCallaghan
Differential Revision: https://reviews.facebook.net/D9651
Summary: Simplified level_ptrs by using a std:vector
Test Plan: make check
Reviewers: sheki, emayanke
Reviewed By: emayanke
CC: leveldb
Differential Revision: https://reviews.facebook.net/D10245
Summary:
FindObsoleteFiles was slow, holding the single big lock, resulted in bad p99 behavior.
Didn't profile anything, but several things could be improved:
1. VersionSet::AddLiveFiles works with std::set, which is by itself slow (a tree).
You also don't know how many dynamic allocations occur just for building up this tree.
switched to std::vector, also added logic to pre-calculate total size and do just one allocation
2. Don't see why env_->GetChildren() needs to be mutex proteced, moved to PurgeObsoleteFiles where
mutex could be unlocked.
3. switched std::set to std:unordered_set, the conversion from vector is also inside PurgeObsoleteFiles
I have a feeling this should pretty much fix it.
Test Plan: make check; db_stress
Reviewers: dhruba, heyongqiang, MarkCallaghan
Reviewed By: dhruba
CC: leveldb, zshao
Differential Revision: https://reviews.facebook.net/D10197
Summary:
This patch allows an application to specify whether to use bufferedio,
reads-via-mmaps and writes-via-mmaps per database. Earlier, there
was a global static variable that was used to configure this functionality.
The default setting remains the same (and is backward compatible):
1. use bufferedio
2. do not use mmaps for reads
3. use mmap for writes
4. use readaheads for reads needed for compaction
I also added a parameter to db_bench to be able to explicitly specify
whether to do readaheads for compactions or not.
Test Plan: make check
Reviewers: sheki, heyongqiang, MarkCallaghan
Reviewed By: sheki
CC: leveldb
Differential Revision: https://reviews.facebook.net/D9429
Summary: Makefile had options to ignore sign-comparisons and unused-parameters, which should be there. Also fixed the specific errors in the code-base
Test Plan: make
Reviewers: chip, dhruba
Reviewed By: dhruba
CC: leveldb
Differential Revision: https://reviews.facebook.net/D9531
Summary:
SizeBeingCompacted was called without any lock protection. This causes
crashes, especially when running db_bench with value_size=128K.
The fix is to compute SizeUnderCompaction while holding the mutex and
passing in these values into the call to Finalize.
(gdb) where
#4 leveldb::VersionSet::SizeBeingCompacted (this=this@entry=0x7f0b490931c0, level=level@entry=4) at db/version_set.cc:1827
#5 0x000000000043a3c8 in leveldb::VersionSet::Finalize (this=this@entry=0x7f0b490931c0, v=v@entry=0x7f0b3b86b480) at db/version_set.cc:1420
#6 0x00000000004418d1 in leveldb::VersionSet::LogAndApply (this=0x7f0b490931c0, edit=0x7f0b3dc8c200, mu=0x7f0b490835b0, new_descriptor_log=<optimized out>) at db/version_set.cc:1016
#7 0x00000000004222b2 in leveldb::DBImpl::InstallCompactionResults (this=this@entry=0x7f0b49083400, compact=compact@entry=0x7f0b2b8330f0) at db/db_impl.cc:1473
#8 0x0000000000426027 in leveldb::DBImpl::DoCompactionWork (this=this@entry=0x7f0b49083400, compact=compact@entry=0x7f0b2b8330f0) at db/db_impl.cc:1757
#9 0x0000000000426690 in leveldb::DBImpl::BackgroundCompaction (this=this@entry=0x7f0b49083400, madeProgress=madeProgress@entry=0x7f0b41bf2d1e, deletion_state=...) at db/db_impl.cc:1268
#10 0x0000000000428f42 in leveldb::DBImpl::BackgroundCall (this=0x7f0b49083400) at db/db_impl.cc:1170
#11 0x000000000045348e in BGThread (this=0x7f0b49023100) at util/env_posix.cc:941
#12 leveldb::(anonymous namespace)::PosixEnv::BGThreadWrapper (arg=0x7f0b49023100) at util/env_posix.cc:874
#13 0x00007f0b4a7cf10d in start_thread (arg=0x7f0b41bf3700) at pthread_create.c:301
#14 0x00007f0b49b4b11d in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:115
Test Plan:
make check
I am running db_bench with a value size of 128K to see if the segfault is fixed.
Reviewers: MarkCallaghan, sheki, emayanke
Reviewed By: sheki
CC: leveldb
Differential Revision: https://reviews.facebook.net/D9279
Summary:
This adds the rate_delay_limit_milliseconds option to make the delay
configurable in MakeRoomForWrite when the max compaction score is too high.
This delay is called the Ln slowdown. This change also counts the Ln slowdown
per level to make it possible to see where the stalls occur.
From IO-bound performance testing, the Level N stalls occur:
* with compression -> at the largest uncompressed level. This makes sense
because compaction for compressed levels is much
slower. When Lx is uncompressed and Lx+1 is compressed
then files pile up at Lx because the (Lx,Lx+1)->Lx+1
compaction process is the first to be slowed by
compression.
* without compression -> at level 1
Task ID: #1832108
Blame Rev:
Test Plan:
run with real data, added test
Revert Plan:
Database Impact:
Memcache Impact:
Other Notes:
EImportant:
- begin *PUBLIC* platform impact section -
Bugzilla: #
- end platform impact -
Reviewers: dhruba
Reviewed By: dhruba
Differential Revision: https://reviews.facebook.net/D9045
Summary:
On some filesystems, pre-allocation can be a considerable
amount of space. xfs in our production environment pre-allocates by
1GB, for instance. By using fallocate to inform the kernel of our
expected file sizes, we eliminate this wasteage (that isn't recovered
until the file is closed which, in the case of LOG files, can be a
considerable amount of time).
Test Plan:
created an xfs loopback filesystem, mounted with
allocsize=4M, and ran db_stress. LOG file without this change was 4M,
and with it it was 128k then grew to normal size.
Reviewers: dhruba
Reviewed By: dhruba
CC: adsharma, leveldb
Differential Revision: https://reviews.facebook.net/D7953
Summary:
Replace manual memory management with std::unique_ptr in a
number of places; not exhaustive, but this fixes a few leaks with file
handles as well as clarifies semantics of the ownership of file handles
with log classes.
Test Plan: db_stress, make check
Reviewers: dhruba
Reviewed By: dhruba
CC: zshao, leveldb, heyongqiang
Differential Revision: https://reviews.facebook.net/D8043
Summary:
Check in LogAndApply if the file size is more than the limit set in
Options.
Things to consider : will this be expensive?
Test Plan: make all check. Inputs on a new unit test?
Reviewers: dhruba
Reviewed By: dhruba
CC: leveldb
Differential Revision: https://reviews.facebook.net/D7701
Summary: Due to how the code handled compactions in Level 0 in `PickCompaction()` it could be the case that two compactions on level 0 ran that produced tables in level 1 that overlap. However, this case seems like it would only occur on a seek compaction which is unlikely on level 0. Furthermore, level 0 and level 1 had to have a certain arrangement of files.
Test Plan:
make check
Reviewers: dhruba, vamsi
Reviewed By: dhruba
CC: leveldb, sheki
Differential Revision: https://reviews.facebook.net/D7923
Summary:
Pretty much a blind copy of the patch in open source.
Hope to get this in before we make a release
Test Plan: make clean check
Reviewers: dhruba, heyongqiang
Reviewed By: dhruba
CC: leveldb
Differential Revision: https://reviews.facebook.net/D7809
Summary:
A compaction is picked based on its score. It is useful to
print the compaction score in the LOG because it aids in
debugging. If one looks at the logs, one can find out why
a compaction was preferred over another.
Test Plan: make clean check
Differential Revision: https://reviews.facebook.net/D7137
Summary:
Scripted and removed all trailing spaces and converted all tabs to
spaces.
Also fixed other lint errors.
All lint errors from this point of time should be taken seriously.
Test Plan: make all check
Reviewers: dhruba
Reviewed By: dhruba
CC: leveldb
Differential Revision: https://reviews.facebook.net/D7059
Summary:
When we expand the range of keys for a level 0 compaction, we
need to invoke ParentFilesInCompaction() only once for the
entire range of keys that is being compacted. We were invoking
it for each file that was being compacted, but this triggers
an assertion because each file's range were contiguous but
non-overlapping.
I renamed ParentFilesInCompaction to ParentRangeInCompaction
to adequately represent that it is the range-of-keys and
not individual files that we compact in a single compaction run.
Here is the assertion that is fixed by this patch.
db_test: db/version_set.cc:585: void leveldb::Version::ExtendOverlappingInputs(int, const leveldb::Slice&, const leveldb::Slice&, std::vector<leveldb::FileMetaData*, std::allocator<leveldb::FileMetaData*> >*, int): Assertion `user_cmp->Compare(flimit, user_begin) >= 0' failed.
Test Plan: make clean check OPT=-g
Reviewers: sheki
Reviewed By: sheki
CC: MarkCallaghan, emayanke, leveldb
Differential Revision: https://reviews.facebook.net/D6963
Summary:
The manifest file contains a series of edits. If the verbose
option is switched on, then print each individual edit in the
manifest file. This helps in debugging.
Test Plan: make clean manifest_dump
Reviewers: emayanke, sheki
Reviewed By: sheki
CC: leveldb
Differential Revision: https://reviews.facebook.net/D6807
Summary:
The manifest file contains a series of edits. If the verbose
option is switched on, then print each individual edit in the
manifest file. This helps in debugging.
Test Plan: make clean manifest_dump
Reviewers: emayanke, sheki
Reviewed By: sheki
CC: leveldb
Differential Revision: https://reviews.facebook.net/D6807
Summary:
When a new version is created, we sort all the files at every
level based on their size. This is necessary because we want
to compact the largest file first. The sorting takes quite a
bit of CPU.
Moved the sorting code to be outside the mutex. Also, the
earlier code was sorting files at all levels but we do not
need to sort the highest-number level because those files
are never the cause of any compaction. To reduce sorting
costs, we sort only the first few files in each level
because it is likely that those are the only files in that
level that will be picked for compaction.
At steady state, I have seen that this patch increase
throughout from 1500 writes/sec to 1700 writes/sec at the
end of a 72 hour run. The cpu saving by not sorting the
last level was not distinctive in this test run because
there were only 100K files in the highest numbered level.
I expect the cpu saving to be significant when the number of
files is much higher.
This is mostly an early preview and not ready for rigorous review.
With this patch, the writs/sec is now bottlenecked not by the sorting code but by GetOverlappingInputs. I am working on a patch to optimize GetOverlappingInputs.
Test Plan: make check
Reviewers: MarkCallaghan, heyongqiang
Reviewed By: heyongqiang
Differential Revision: https://reviews.facebook.net/D6411
Summary:
The Version::GetOverlappingInputs() is called multiple times in
the compaction code path. Eack invocation does a binary search
for overlapping files in the specified key range.
This patch remembers the offset of an overlapped file when
GetOverlappingInputs() is called the first time within
a compaction run. Suceeding calls to GetOverlappingInputs()
uses the remembered index to avoid the binary search.
I measured that 1000 iterations of GetOverlappingInputs
takes around 4500 microseconds without this patch. If I use
this patch with the hint on every invocation, then 1000
iterations take about 3900 microsecond.
Test Plan: make check OPT=-g
Reviewers: heyongqiang
Reviewed By: heyongqiang
CC: MarkCallaghan, emayanke, sheki
Differential Revision: https://reviews.facebook.net/D6513
Summary:
The method Version::GetOverlappingInputs used a sequential search
to map a kay-range to a set of files. But the files are arranged
in ascending order of key, so a biary search is more effective.
This patch implements Version::GetOverlappingInputsBinarySearch
that finds one file that corresponds to the specified key range
and then iterates backwards and forwards to find all overlapping
files.
This patch is critical for making compactions efficient, especially
when there are thousands of files in a single level.
I measured that 1000 iterations of TEST_MaxNextLevelOverlappingBytes
takes 16000 microseconds without this patch. With this patch, the
same method takes about 4600 microseconds.
Test Plan: Almost all unit tests in db_test uses this method to lookup keys.
Reviewers: heyongqiang
Reviewed By: heyongqiang
CC: MarkCallaghan, emayanke, sheki
Differential Revision: https://reviews.facebook.net/D6465
Summary: as subject.
Test Plan: manually test it, will add a testcase
Reviewers: dhruba, MarkCallaghan
Differential Revision: https://reviews.facebook.net/D6345
Summary:
Adds a method that returns the score for the next level that most
needs compaction. That method is then used by db_bench to rate limit threads.
Threads are put to sleep at the end of each stats interval until the score
is less than the limit. The limit is set via the --rate_limit=$double option.
The specified value must be > 1.0. Also adds the option --stats_per_interval
to enable additional metrics reported every stats interval.
Task ID: #
Blame Rev:
Test Plan:
run db_bench
Revert Plan:
Database Impact:
Memcache Impact:
Other Notes:
EImportant:
- begin *PUBLIC* platform impact section -
Bugzilla: #
- end platform impact -
Reviewers: dhruba
Reviewed By: dhruba
Differential Revision: https://reviews.facebook.net/D6243
Summary:
It is best if we pick the largest file to compact in a level.
This reduces the write amplification factor for compactions.
Each level has an auxiliary data structure called files_by_size_
that sorts all files by their size. This data structure is
updated when a new version is created.
Test Plan: make check
Differential Revision: https://reviews.facebook.net/D6195
published in https://reviews.facebook.net/D5997.
Summary:
This patch allows compaction to occur in multiple background threads
concurrently.
If a manual compaction is issued, the system falls back to a
single-compaction-thread model. This is done to ensure correctess
and simplicity of code. When the manual compaction is finished,
the system resumes its concurrent-compaction mode automatically.
The updates to the manifest are done via group-commit approach.
Test Plan: run db_bench
Summary:
The GetLiveFiles() api lists the set of sst files and the current
MANIFEST file. But the database continues to append new data to the
MANIFEST file even when the application is backing it up to the
backup location. This means that the database-version that is
stored in the MANIFEST FILE in the backup location
does not correspond to the sst files returned by GetLiveFiles.
This API adds a new parameter to GetLiveFiles. This new parmeter
returns the current size of the MANIFEST file.
Test Plan: Unit test attached.
Reviewers: heyongqiang
Reviewed By: heyongqiang
Differential Revision: https://reviews.facebook.net/D5631
Summary:
A set of apis that allows an application to backup data from the
leveldb database based on a set of files.
Test Plan: unint test attached. more coming soon.
Reviewers: heyongqiang
Reviewed By: heyongqiang
Differential Revision: https://reviews.facebook.net/D5439
Summary:
Clean up compiler warnings generated by -Wall option.
make clean all OPT=-Wall
This is a pre-requisite before making a new release.
Test Plan: compile and run unit tests
Reviewers: heyongqiang
Reviewed By: heyongqiang
Differential Revision: https://reviews.facebook.net/D5019
Summary:
as subject.
A new log is written to scribe via thrift client when a new db is opened and when there is
a compaction.
a new option var scribe_log_db_stats is added.
Test Plan: manually checked using command "ptail -time 0 leveldb_deploy_stats"
Reviewers: dhruba
Differential Revision: https://reviews.facebook.net/D4659
- Replace raw slice comparison with a call to user comparator.
Added test for custom comparators.
- Fix end of namespace comments.
- Fixed bug in picking inputs for a level-0 compaction.
When finding overlapping files, the covered range may expand
as files are added to the input set. We now correctly expand
the range when this happens instead of continuing to use the
old range. For example, suppose L0 contains files with the
following ranges:
F1: a .. d
F2: c .. g
F3: f .. j
and the initial compaction target is F3. We used to search
for range f..j which yielded {F2,F3}. However we now expand
the range as soon as another file is added. In this case,
when F2 is added, we expand the range to c..j and restart the
search. That picks up file F1 as well.
This change fixes a bug related to deleted keys showing up
incorrectly after a compaction as described in Issue 44.
(Sync with upstream @25072954)
- Added DB::CompactRange() method.
Changed manual compaction code so it breaks up compactions of
big ranges into smaller compactions.
Changed the code that pushes the output of memtable compactions
to higher levels to obey the grandparent constraint: i.e., we
must never have a single file in level L that overlaps too
much data in level L+1 (to avoid very expensive L-1 compactions).
Added code to pretty-print internal keys.
- Fixed bug where we would not detect overlap with files in
level-0 because we were incorrectly using binary search
on an array of files with overlapping ranges.
Added "leveldb.sstables" property that can be used to dump
all of the sstables and ranges that make up the db state.
- Removing post_write_snapshot support. Email to leveldb mailing
list brought up no users, just confusion from one person about
what it meant.
- Fixing static_cast char to unsigned on BIG_ENDIAN platforms.
Fixes Issue 35 and Issue 36.
- Comment clarification to address leveldb Issue 37.
- Change license in posix_logger.h to match other files.
- A build problem where uint32 was used instead of uint32_t.
Sync with upstream @24408625
- Fix bug in Get: when it triggers a compaction, it could sometimes
mark the compaction with the wrong level (if there was a gap
in the set of levels examined for the Get).
- Do not hold mutex while writing to the log file or to the
MANIFEST file.
Added a new benchmark that runs a writer thread concurrently with
reader threads.
Percentiles
------------------------------
micros/op: avg median 99 99.9 99.99 99.999 max
------------------------------------------------------
before: 42 38 110 225 32000 42000 48000
after: 24 20 55 65 130 1100 7000
- Fixed race in optimized Get. It should have been using the
pinned memtables, not the current memtables.
git-svn-id: https://leveldb.googlecode.com/svn/trunk@50 62dab493-f737-651d-591e-8d6aee1b9529
Slight tweak to the no-overlap optimization: only push to
level 2 to reduce the amount of wasted space when the same
small key range is being repeatedly overwritten.
Fix for Issue 18: Avoid failure on Windows by avoiding
deletion of lock file until the end of DestroyDB().
Fix for Issue 19: Disregard sequence numbers when checking for
overlap in sstable ranges. This fixes issue 19: when writing
the same key over and over again, we would generate a sequence
of sstables that were never merged together since their sequence
numbers were disjoint.
Don't ignore map/unmap error checks.
Miscellaneous fixes for small problems Sanjay found while diagnosing
issue/9 and issue/16 (corruption_testr failures).
- log::Reader reports the record type when it finds an unexpected type.
- log::Reader no longer reports an error when it encounters an expected
zero record regardless of the setting of the "checksum" flag.
- Added a missing forward declaration.
- Documented a side-effects of larger write buffer sizes
(longer recovery time).
git-svn-id: https://leveldb.googlecode.com/svn/trunk@37 62dab493-f737-651d-591e-8d6aee1b9529
- Implemented Get() directly instead of building on top of a full
merging iterator stack. This speeds up the "readrandom" benchmark
by up to 15-30%.
- Fixed an opensource compilation problem.
Added --db=<name> flag to control where the database is placed.
- Automatically compact a file when we have done enough
overlapping seeks to that file.
- Fixed a performance bug where we would read from at least one
file in a level even if none of the files overlapped the key
being read.
- Makefile fix for Mac OSX installations that have XCode 4 without XCode 3.
- Unified the two occurrences of binary search in a file-list
into one routine.
- Found and fixed a bug where we would unnecessarily search the
last file when looking for a key larger than all data in the
level.
- A fix to avoid the need for trivial move compactions and
therefore gets rid of two out of five syncs in "fillseq".
- Removed the MANIFEST file write when switching to a new
memtable/log-file for a 10-20% improvement on fill speed on ext4.
- Adding a SNAPPY setting in the Makefile for folks who have
Snappy installed. Snappy compresses values and speeds up writes.
git-svn-id: https://leveldb.googlecode.com/svn/trunk@32 62dab493-f737-651d-591e-8d6aee1b9529
* env_chromium.cc should not export symbols.
* Fix MSVC warnings.
* Removed large value support.
* Fix broken reference to documentation file
git-svn-id: https://leveldb.googlecode.com/svn/trunk@24 62dab493-f737-651d-591e-8d6aee1b9529
Dhruba Borthakur