Mercurial > hg
view tests/test-verify.t @ 29830:92ac2baaea86
revlog: use an LRU cache for delta chain bases
Profiling using statprof revealed a hotspot during changegroup
application calculating delta chain bases on generaldelta repos.
Essentially, revlog._addrevision() was performing a lot of redundant
work tracing the delta chain as part of determining when the chain
distance was acceptable. This was most pronounced when adding
revisions to manifests, which can have delta chains thousands of
revisions long.
There was a delta chain base cache on revlogs before, but it only
captured a single revision. This was acceptable before generaldelta,
when _addrevision would build deltas from the previous revision and
thus we'd pretty much guarantee a cache hit when resolving the delta
chain base on a subsequent _addrevision call. However, it isn't
suitable for generaldelta because parent revisions aren't necessarily
the last processed revision.
This patch converts the delta chain base cache to an LRU dict cache.
The cache can hold multiple entries, so generaldelta repos have a
higher chance of getting a cache hit.
The impact of this change when processing changegroup additions is
significant. On a generaldelta conversion of the "mozilla-unified"
repo (which contains heads of the main Firefox repositories in
chronological order - this means there are lots of transitions between
heads in revlog order), this change has the following impact when
performing an `hg unbundle` of an uncompressed bundle of the repo:
before: 5:42 CPU time
after: 4:34 CPU time
Most of this time is saved when applying the changelog and manifest
revlogs:
before: 2:30 CPU time
after: 1:17 CPU time
That nearly a 50% reduction in CPU time applying changesets and
manifests!
Applying a gzipped bundle of the same repo (effectively simulating a
`hg clone` over HTTP) showed a similar speedup:
before: 5:53 CPU time
after: 4:46 CPU time
Wall time improvements were basically the same as CPU time.
I didn't measure explicitly, but it feels like most of the time
is saved when processing manifests. This makes sense, as large
manifests tend to have very long delta chains and thus benefit the
most from this cache.
So, this change effectively makes changegroup application (which is
used by `hg unbundle`, `hg clone`, `hg pull`, `hg unshelve`, and
various other commands) significantly faster when delta chains are
long (which can happen on repos with large numbers of files and thus
large manifests).
In theory, this change can result in more memory utilization. However,
we're caching a dict of ints. At most we have 200 ints + Python object
overhead per revlog. And, the cache is really only populated when
performing read-heavy operations, such as adding changegroups or
scanning an individual revlog. For memory bloat to be an issue, we'd
need to scan/read several revisions from several revlogs all while
having active references to several revlogs. I don't think there are
many operations that do this, so I don't think memory bloat from the
cache will be an issue.
author | Gregory Szorc <gregory.szorc@gmail.com> |
---|---|
date | Mon, 22 Aug 2016 21:48:50 -0700 |
parents | 408446e4b10c |
children | c059286a0f9c |
line wrap: on
line source
prepare repo $ hg init a $ cd a $ echo "some text" > FOO.txt $ echo "another text" > bar.txt $ echo "more text" > QUICK.txt $ hg add adding FOO.txt adding QUICK.txt adding bar.txt $ hg ci -mtest1 verify $ hg verify checking changesets checking manifests crosschecking files in changesets and manifests checking files 3 files, 1 changesets, 3 total revisions verify with journal $ touch .hg/store/journal $ hg verify abandoned transaction found - run hg recover checking changesets checking manifests crosschecking files in changesets and manifests checking files 3 files, 1 changesets, 3 total revisions $ rm .hg/store/journal introduce some bugs in repo $ cd .hg/store/data $ mv _f_o_o.txt.i X_f_o_o.txt.i $ mv bar.txt.i xbar.txt.i $ rm _q_u_i_c_k.txt.i $ hg verify checking changesets checking manifests crosschecking files in changesets and manifests checking files warning: revlog 'data/FOO.txt.i' not in fncache! 0: empty or missing FOO.txt FOO.txt@0: manifest refers to unknown revision f62022d3d590 warning: revlog 'data/QUICK.txt.i' not in fncache! 0: empty or missing QUICK.txt QUICK.txt@0: manifest refers to unknown revision 88b857db8eba warning: revlog 'data/bar.txt.i' not in fncache! 0: empty or missing bar.txt bar.txt@0: manifest refers to unknown revision 256559129457 3 files, 1 changesets, 0 total revisions 3 warnings encountered! hint: run "hg debugrebuildfncache" to recover from corrupt fncache 6 integrity errors encountered! (first damaged changeset appears to be 0) [1] $ cd ../../.. $ cd .. Set up a repo for testing missing revlog entries $ hg init missing-entries $ cd missing-entries $ echo 0 > file $ hg ci -Aqm0 $ cp -r .hg/store .hg/store-partial $ echo 1 > file $ hg ci -Aqm1 $ cp -r .hg/store .hg/store-full Entire changelog missing $ rm .hg/store/00changelog.* $ hg verify -q 0: empty or missing changelog manifest@0: d0b6632564d4 not in changesets manifest@1: 941fc4534185 not in changesets 3 integrity errors encountered! (first damaged changeset appears to be 0) [1] $ cp -r .hg/store-full/. .hg/store Entire manifest log missing $ rm .hg/store/00manifest.* $ hg verify -q 0: empty or missing manifest 1 integrity errors encountered! (first damaged changeset appears to be 0) [1] $ cp -r .hg/store-full/. .hg/store Entire filelog missing $ rm .hg/store/data/file.* $ hg verify -q warning: revlog 'data/file.i' not in fncache! 0: empty or missing file file@0: manifest refers to unknown revision 362fef284ce2 file@1: manifest refers to unknown revision c10f2164107d 1 warnings encountered! hint: run "hg debugrebuildfncache" to recover from corrupt fncache 3 integrity errors encountered! (first damaged changeset appears to be 0) [1] $ cp -r .hg/store-full/. .hg/store Entire changelog and manifest log missing $ rm .hg/store/00changelog.* $ rm .hg/store/00manifest.* $ hg verify -q warning: orphan revlog 'data/file.i' 1 warnings encountered! $ cp -r .hg/store-full/. .hg/store Entire changelog and filelog missing $ rm .hg/store/00changelog.* $ rm .hg/store/data/file.* $ hg verify -q 0: empty or missing changelog manifest@0: d0b6632564d4 not in changesets manifest@1: 941fc4534185 not in changesets warning: revlog 'data/file.i' not in fncache! ?: empty or missing file file@0: manifest refers to unknown revision 362fef284ce2 file@1: manifest refers to unknown revision c10f2164107d 1 warnings encountered! hint: run "hg debugrebuildfncache" to recover from corrupt fncache 6 integrity errors encountered! (first damaged changeset appears to be 0) [1] $ cp -r .hg/store-full/. .hg/store Entire manifest log and filelog missing $ rm .hg/store/00manifest.* $ rm .hg/store/data/file.* $ hg verify -q 0: empty or missing manifest warning: revlog 'data/file.i' not in fncache! 0: empty or missing file 1 warnings encountered! hint: run "hg debugrebuildfncache" to recover from corrupt fncache 2 integrity errors encountered! (first damaged changeset appears to be 0) [1] $ cp -r .hg/store-full/. .hg/store Changelog missing entry $ cp -f .hg/store-partial/00changelog.* .hg/store $ hg verify -q manifest@?: rev 1 points to nonexistent changeset 1 manifest@?: 941fc4534185 not in changesets file@?: rev 1 points to nonexistent changeset 1 (expected 0) 1 warnings encountered! 3 integrity errors encountered! [1] $ cp -r .hg/store-full/. .hg/store Manifest log missing entry $ cp -f .hg/store-partial/00manifest.* .hg/store $ hg verify -q manifest@1: changeset refers to unknown revision 941fc4534185 file@1: c10f2164107d not in manifests 2 integrity errors encountered! (first damaged changeset appears to be 1) [1] $ cp -r .hg/store-full/. .hg/store Filelog missing entry $ cp -f .hg/store-partial/data/file.* .hg/store/data $ hg verify -q file@1: manifest refers to unknown revision c10f2164107d 1 integrity errors encountered! (first damaged changeset appears to be 1) [1] $ cp -r .hg/store-full/. .hg/store Changelog and manifest log missing entry $ cp -f .hg/store-partial/00changelog.* .hg/store $ cp -f .hg/store-partial/00manifest.* .hg/store $ hg verify -q file@?: rev 1 points to nonexistent changeset 1 (expected 0) file@?: c10f2164107d not in manifests 1 warnings encountered! 2 integrity errors encountered! [1] $ cp -r .hg/store-full/. .hg/store Changelog and filelog missing entry $ cp -f .hg/store-partial/00changelog.* .hg/store $ cp -f .hg/store-partial/data/file.* .hg/store/data $ hg verify -q manifest@?: rev 1 points to nonexistent changeset 1 manifest@?: 941fc4534185 not in changesets file@?: manifest refers to unknown revision c10f2164107d 3 integrity errors encountered! [1] $ cp -r .hg/store-full/. .hg/store Manifest and filelog missing entry $ cp -f .hg/store-partial/00manifest.* .hg/store $ cp -f .hg/store-partial/data/file.* .hg/store/data $ hg verify -q manifest@1: changeset refers to unknown revision 941fc4534185 1 integrity errors encountered! (first damaged changeset appears to be 1) [1] $ cp -r .hg/store-full/. .hg/store Corrupt changelog base node to cause failure to read revision $ printf abcd | dd conv=notrunc of=.hg/store/00changelog.i bs=1 seek=16 \ > 2> /dev/null $ hg verify -q 0: unpacking changeset 08b1860757c2: * (glob) manifest@?: rev 0 points to unexpected changeset 0 manifest@?: d0b6632564d4 not in changesets file@?: rev 0 points to unexpected changeset 0 (expected 1) 1 warnings encountered! 4 integrity errors encountered! (first damaged changeset appears to be 0) [1] $ cp -r .hg/store-full/. .hg/store Corrupt manifest log base node to cause failure to read revision $ printf abcd | dd conv=notrunc of=.hg/store/00manifest.i bs=1 seek=16 \ > 2> /dev/null $ hg verify -q manifest@0: reading delta d0b6632564d4: * (glob) file@0: 362fef284ce2 not in manifests 2 integrity errors encountered! (first damaged changeset appears to be 0) [1] $ cp -r .hg/store-full/. .hg/store Corrupt filelog base node to cause failure to read revision $ printf abcd | dd conv=notrunc of=.hg/store/data/file.i bs=1 seek=16 \ > 2> /dev/null $ hg verify -q file@0: unpacking 362fef284ce2: * (glob) 1 integrity errors encountered! (first damaged changeset appears to be 0) [1] $ cp -r .hg/store-full/. .hg/store $ cd .. test changelog without a manifest $ hg init b $ cd b $ hg branch foo marked working directory as branch foo (branches are permanent and global, did you want a bookmark?) $ hg ci -m branchfoo $ hg verify checking changesets checking manifests crosschecking files in changesets and manifests checking files 0 files, 1 changesets, 0 total revisions test revlog corruption $ touch a $ hg add a $ hg ci -m a $ echo 'corrupted' > b $ dd if=.hg/store/data/a.i of=start bs=1 count=20 2>/dev/null $ cat start b > .hg/store/data/a.i $ hg verify checking changesets checking manifests crosschecking files in changesets and manifests checking files a@1: broken revlog! (index data/a.i is corrupted) warning: orphan revlog 'data/a.i' 1 files, 2 changesets, 0 total revisions 1 warnings encountered! 1 integrity errors encountered! (first damaged changeset appears to be 1) [1] $ cd .. test revlog format 0 $ revlog-formatv0.py $ cd formatv0 $ hg verify repository uses revlog format 0 checking changesets checking manifests crosschecking files in changesets and manifests checking files 1 files, 1 changesets, 1 total revisions $ cd ..