Summary:
LevelDB should delete almost-new keys when a long-open snapshot exists.
The previous behavior is to keep all versions that were created after the
oldest open snapshot. This can lead to database size bloat for
high-update workloads when there are long-open snapshots and long-open
snapshot will be used for logical backup. By "almost new" I mean that the
key was updated more than once after the oldest snapshot.
If there were two snapshots with seq numbers s1 and s2 (s1 < s2), and if
we find two instances of the same key k1 that lie entirely within s1 and
s2 (i.e. s1 < k1 < s2), then the earlier version
of k1 can be safely deleted because that version is not visible in any snapshot.
Test Plan:
unit test attached
make clean check
Differential Revision: https://reviews.facebook.net/D6999
Summary:
Print out status at the end of a compaction run. This helps in
debugging.
Test Plan: make clean check
Reviewers: sheki
Reviewed By: sheki
Differential Revision: https://reviews.facebook.net/D7035
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:
On fast filesystems (e.g. /dev/shm and ext4), the flushing
of memstore to disk was fast and quick, and the background compaction
thread was not getting scheduled fast enough to delete obsolete
files before the db was closed. This caused the repair method
to pick up those files that were not part of the db and the unit
test was failing.
The fix is to enhance the unti test to run a compaction before
closing the database so that all files that are not part of the
database are truly deleted from the filesystem.
Test Plan: make c_test; ./c_test
Reviewers: chip, emayanke, sheki
Reviewed By: chip
CC: leveldb
Differential Revision: https://reviews.facebook.net/D6915
Summary:
It would appear our unit tests make use of code from ldb_cmd,
and don't always require a valid database handle. D6855 was not aware
db_ could sometimes be NULL for such commands, and so it broke
reduce_levels_test.
This moves the check elsewhere to (at least) fix the 'ldb dump' case of
segfaulting when it couldn't open a database.
Test Plan: make check
Reviewers: dhruba
Reviewed By: dhruba
CC: leveldb
Differential Revision: https://reviews.facebook.net/D6903
Summary:
Link statically against snappy, using the gvfs one for facebook
environments, and the bundled one otherwise.
In addition, fix a few minor segfaults in ldb when it couldn't open the
database, and update .gitignore to include a few other build artifacts.
Test Plan: make check
Reviewers: dhruba
Reviewed By: dhruba
CC: leveldb
Differential Revision: https://reviews.facebook.net/D6855
Summary:
The compaction process takes some files from LevelK and
merges it into LevelK+1. The number of files it picks from
LevelK was capped such a way that the total amount of
data picked does not exceed the maxfilesize of that level.
This essentially meant that only one file from LevelK
is picked for a single compaction.
For bulkloads, we would like to take many many file from
LevelK and compact them using a single compaction run.
This patch introduces a option called the 'source_compaction_factor'
(similar to expanded_compaction_factor). It is a multiplier
that is multiplied by the maxfilesize of that level to arrive
at the limit that is used to throttle the number of source
files from LevelK. For bulk loads, set source_compaction_factor
to a very high number so that multiple files from the same
level are picked for compaction in a single compaction.
The default value of source_compaction_factor is 1, so that
we can keep backward compatibilty with existing compaction semantics.
Test Plan: make clean check
Reviewers: emayanke, sheki
Reviewed By: emayanke
CC: leveldb
Differential Revision: https://reviews.facebook.net/D6867
Summary:
This option is needed for fast bulk uploads. The goal is to load
all the data into files in L0 without any interference from
background compactions.
Test Plan: make clean check
Reviewers: sheki
Reviewed By: sheki
CC: leveldb
Differential Revision: https://reviews.facebook.net/D6849
Summary:
The method Finalize() recomputes the compaction score of each
level and then sorts these score from largest to smallest. The
idea is that the level with the largest compaction score will
be a better candidate for compaction. There are usually very
few levels, and a bubble sort code was used to sort these
compaction scores. There existed a bug in the sorting code that
skipped looking at the score for the n-1 level. This meant that
even if the compaction score of the n-1 level is large, it will
not be picked for compaction.
This patch fixes the bug and also introduces "asserts" in the
code to detect any possible inconsistencies caused by future bugs.
This bug existed in the very first code change that introduced
multi-threaded compaction to the leveldb code. That version of
code was committed on Oct 19th via
1ca0584345
Test Plan: make clean check OPT=-g
Reviewers: emayanke, sheki, MarkCallaghan
Reviewed By: sheki
CC: leveldb
Differential Revision: https://reviews.facebook.net/D6837
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:
StringStream.clear() does not clear the stream. It sets some flags.
Who knew? Fixing that is not printing the stuff again and again.
Test Plan: ran it on a local db
Reviewers: dhruba, emayanke
Reviewed By: dhruba
Differential Revision: https://reviews.facebook.net/D6795
Summary: The new function MinLevelToCompress in db_test.cc was incomplete. It needs to tell the calling function-TEST whether the test has to be skipped or not
Test Plan: make all;./db_test
Reviewers: dhruba, heyongqiang
Reviewed By: dhruba
CC: sheki
Differential Revision: https://reviews.facebook.net/D6771
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:
StringStream.clear() does not clear the stream. It sets some flags.
Who knew? Fixing that is not printing the stuff again and again.
Test Plan: ran it on a local db
Reviewers: dhruba, emayanke
Reviewed By: dhruba
Differential Revision: https://reviews.facebook.net/D6795
Summary:
dbstress has an option to reopen the database. Make it such that the
previous handle is not closed before we reopen, this simulates a
situation similar to a process crash.
Added new api to DMImpl to remove the lock file.
Test Plan: run db_stress
Reviewers: emayanke
Reviewed By: emayanke
CC: leveldb
Differential Revision: https://reviews.facebook.net/D6777
Summary: The new function MinLevelToCompress in db_test.cc was incomplete. It needs to tell the calling function-TEST whether the test has to be skipped or not
Test Plan: make all;./db_test
Reviewers: dhruba, heyongqiang
Reviewed By: dhruba
CC: sheki
Differential Revision: https://reviews.facebook.net/D6771
Summary:
Create more than one background compaction thread if specified.
This code peice is similar to what exists in db_bench.
Test Plan: make check
Differential Revision: https://reviews.facebook.net/D6753
Summary:
The value specified in max_grandparent_overlap_factor is used to
limit the file size in a compaction run. This patch makes it
configurable when using db_bench.
Test Plan: make clean db_bench
Reviewers: MarkCallaghan, heyongqiang
Reviewed By: heyongqiang
CC: leveldb
Differential Revision: https://reviews.facebook.net/D6729
Summary:
The fbcode compilation was always switching on msse by default.
This patch keeps the same behaviour but allows the compilation
process to switch off msse if needed.
If one does not want to use sse, then do the following:
export USE_SSE=0
make clean all
Test Plan: make clean all
Reviewers: heyongqiang
Reviewed By: heyongqiang
CC: leveldb
Differential Revision: https://reviews.facebook.net/D6717
Summary:
There are applications that operate on multiple leveldb instances.
These applications will like to pass in an opaque type for each
leveldb instance and this type should be passed back to the application
with every invocation of the CompactionFilter api.
Test Plan: Enehanced unit test for opaque parameter to CompactionFilter.
Reviewers: heyongqiang
Reviewed By: heyongqiang
CC: MarkCallaghan, sheki, emayanke
Differential Revision: https://reviews.facebook.net/D6711
Summary:
Compilation used to fail with the error:
db/version_set.cc:1773: error: ‘number_of_files_to_sort_’ is not a member of ‘leveldb::VersionSet’
I created a new method called CheckConsistencyForDeletes() so that
all the high cost checking is done only when OPT=-g is specified.
I also fixed a bug in PickCompactionBySize that was triggered when
OPT=-g was switched on. The base_index in the compaction record
was not set correctly.
Test Plan: make check OPT=-g
Differential Revision: https://reviews.facebook.net/D6687
Summary:
The db_bench utility was broken in 1.5.4.fb because of a
signed-unsigned comparision.
The static variable FLAGS_min_level_to_compress was recently
changed from int to 'unsigned in' but it is initilized to a
nagative value -1.
The segfault is of this type:
Program received signal SIGSEGV, Segmentation fault.
Open (this=0x7fffffffdee0) at db/db_bench.cc:939
939 db/db_bench.cc: No such file or directory.
(gdb) where
Test Plan: run db_bench with no options.
Reviewers: heyongqiang
Reviewed By: heyongqiang
CC: MarkCallaghan, emayanke, sheki
Differential Revision: https://reviews.facebook.net/D6663
Summary: FLAGS_reopen (configurable) specifies the number of times the databse is to be reopened. FLAGS_ops_per_thread is divided into points based on that reopen field. At these points all threads come together to wait for the databse to reopen. Each thread "votes" for the database to reopen and when all have voted, the database reopens.
Test Plan: make all;./db_stress
Reviewers: dhruba, MarkCallaghan, sheki, asad, heyongqiang
Reviewed By: dhruba
Differential Revision: https://reviews.facebook.net/D6627
Summary:
I changed the reduce_num_levels logic to avoid "compactRange()" call if the current number of levels in use (levels that contain files) is smaller than the new num of levels.
And that change breaks the assert in reduce_levels_test
Test Plan: run reduce_levels_test
Reviewers: dhruba, MarkCallaghan
Reviewed By: dhruba
CC: emayanke, sheki
Differential Revision: https://reviews.facebook.net/D6651
Summary:
make clean check OPT=-g fails
leveldb::DBStatistics::getTickerCount(leveldb::Tickers)’:
./db/db_statistics.h:34: error: ‘MAX_NO_TICKERS’ was not declared in this scope
util/ldb_cmd.cc:255: warning: left shift count >= width of type
Test Plan:
make clean check OPT=-g
Reviewers:
CC:
Task ID: #
Blame Rev:
Summary: Record BloomFliter hits and drop off reasons during compaction.
Test Plan: Unit tests work.
Reviewers: dhruba, heyongqiang
Reviewed By: dhruba
Differential Revision: https://reviews.facebook.net/D6591
Summary:
disable size compaction in ldb reduce_levels, this will avoid compactions rather than the manual comapction,
added --compression=none|snappy|zlib|bzip2 and --file_size= per-file size to ldb reduce_levels command
Test Plan: run ldb
Reviewers: dhruba, MarkCallaghan
Reviewed By: dhruba
CC: sheki, emayanke
Differential Revision: https://reviews.facebook.net/D6597
Summary: Stress test modified to do deletes and later verify them
Test Plan: running the test: db_stress
Reviewers: dhruba, heyongqiang, asad, sheki, MarkCallaghan
Reviewed By: dhruba
Differential Revision: https://reviews.facebook.net/D6567
Summary:
Prototype stat's collection. Diff is a good estimate of what
the final code will look like.
A few assumptions :
* Used a global static instance of the statistics object. Plan to pass
it to each internal function. Static allows metrics only at app
level.
* In the Ticker's do not do any locking. Depend on the mutex at each
function of LevelDB. If we ever remove the mutex, we should change
here too. The other option is use atomic objects anyways as there
won't be any contention as they will be always acquired only by one
thread.
* The counters are dumb, increment through lifecycle. Plan to use ods
etc to get last5min stat etc.
Test Plan:
made changes in db_bench
Ran ./db_bench --statistics=1 --num=10000 --cache_size=5000
This will print the cache hit/miss stats.
Reviewers: dhruba, heyongqiang
Differential Revision: https://reviews.facebook.net/D6441
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:
Added a conditional flush in ~DBImpl to flush.
There is still a chance of writes not being persisted if there is a
crash (not a clean shutdown) before the DBImpl instance is destroyed.
Test Plan: modified db_test to meet the new expectations.
Reviewers: dhruba, heyongqiang
Differential Revision: https://reviews.facebook.net/D6519
Summary:
The default compilation process now uses "-Wall" to compile.
Fix all compilation error generated by gcc.
Test Plan: make all check
Reviewers: heyongqiang, emayanke, sheki
Reviewed By: heyongqiang
CC: MarkCallaghan
Differential Revision: https://reviews.facebook.net/D6525
