lfs: fix the stall and corruption issue when concurrently uploading blobs
We've avoided the issue up to this point by gating worker usage with an
experimental config. See 10e62d5efa73, and the thread linked there for some of
the initial diagnosis, but essentially some data was being read from the blob
before an error occurred and `keepalive` retried, but didn't rewind the file
pointer. So the leading data was lost from the blob on the server, and the
connection stalled, trying to send more data than available.
In trying to recreate this, I was unable to do so uploading from Windows to
CentOS 7. But it reproduced every time going from CentOS 7 to another CentOS 7
over https.
I found recent fixes in the FaceBook repo to address this[1][2]. The commit
message for the first is:
The KeepAlive HTTP implementation is bugged in it's retry logic, it supports
reading from a file pointer, but doesn't support rewinding of the seek cursor
when it performs a retry. So it can happen that an upload fails for whatever
reason and will then 'hang' on the retry event.
The sequence of events that get triggered are:
- Upload file A, goes OK. Keep-Alive caches connection.
- Upload file B, fails due to (for example) failing Keep-Alive, but LFS file
pointer has been consumed for the upload and fd has been closed.
- Retry for file B starts, sets the Content-Length properly to the expected
file size, but since file pointer has been consumed no data will be uploaded,
causing the server to wait for the uploaded data until either client or
server reaches a timeout, making it seem as our mercurial process hangs.
This is just a stop-gap measure to prevent this behavior from blocking Mercurial
(LFS has retry logic). A proper solutions need to be build on top of this
stop-gap measure: for upload from file pointers, we should support fseek() on
the interface. Since we expect to consume the whole file always anyways, this
should be safe. This way we can seek back to the beginning on a retry.
I ported those two patches, and it works. But I see that `url._sendfile()` does
a rewind on `httpsendfile` objects[3], so maybe it's better to keep this all in
one place and avoid a second seek. We may still want the first FaceBook patch
as extra protection for this problem in general. The other two uses of
`httpsendfile` are in the wire protocol to upload bundles, and to upload
largefiles. Neither of these appear to use a worker, and I'm not sure why
workers seem to trigger this, or if this could have happened without a worker.
Since `httpsendfile` already has a `close()` method, that is dropped. That
class also explicitly says there's no `__len__` attribute, so that is removed
too. The override for `read()` is necessary to avoid the progressbar usage per
file.
[1] https://github.com/facebookexperimental/eden/commit/c350d6536d90c044c837abdd3675185644481469
[2] https://github.com/facebookexperimental/eden/commit/77f0d3fd0415e81b63e317e457af9c55c46103ee
[3] https://www.mercurial-scm.org/repo/hg/file/5.2.2/mercurial/url.py#l176
Differential Revision: https://phab.mercurial-scm.org/D7962
#require no-reposimplestore
Create an ondisk bundlestore in .hg/scratchbranches
$ . "$TESTDIR/library-infinitepush.sh"
$ cp $HGRCPATH $TESTTMP/defaulthgrc
$ setupcommon
$ mkcommit() {
> echo "$1" > "$1"
> hg add "$1"
> hg ci -m "$1"
> }
$ hg init repo
$ cd repo
Check that we can send a scratch on the server and it does not show there in
the history but is stored on disk
$ setupserver
$ cd ..
$ hg clone ssh://user@dummy/repo client -q
$ cd client
$ mkcommit initialcommit
$ hg push -r .
pushing to ssh://user@dummy/repo
searching for changes
remote: adding changesets
remote: adding manifests
remote: adding file changes
remote: added 1 changesets with 1 changes to 1 files
$ mkcommit scratchcommit
$ hg push -r . -B scratch/mybranch
pushing to ssh://user@dummy/repo
searching for changes
remote: pushing 1 commit:
remote: 20759b6926ce scratchcommit
$ hg log -G
@ changeset: 1:20759b6926ce
| bookmark: scratch/mybranch
| tag: tip
| user: test
| date: Thu Jan 01 00:00:00 1970 +0000
| summary: scratchcommit
|
o changeset: 0:67145f466344
user: test
date: Thu Jan 01 00:00:00 1970 +0000
summary: initialcommit
$ hg log -G -R ../repo
o changeset: 0:67145f466344
tag: tip
user: test
date: Thu Jan 01 00:00:00 1970 +0000
summary: initialcommit
$ find ../repo/.hg/scratchbranches | sort
../repo/.hg/scratchbranches
../repo/.hg/scratchbranches/filebundlestore
../repo/.hg/scratchbranches/filebundlestore/b9
../repo/.hg/scratchbranches/filebundlestore/b9/e1
../repo/.hg/scratchbranches/filebundlestore/b9/e1/b9e1ee5f93fb6d7c42496fc176c09839639dd9cc
../repo/.hg/scratchbranches/index
../repo/.hg/scratchbranches/index/bookmarkmap
../repo/.hg/scratchbranches/index/bookmarkmap/scratch
../repo/.hg/scratchbranches/index/bookmarkmap/scratch/mybranch
../repo/.hg/scratchbranches/index/nodemap
../repo/.hg/scratchbranches/index/nodemap/20759b6926ce827d5a8c73eb1fa9726d6f7defb2
From another client we can get the scratchbranch if we ask for it explicitely
$ cd ..
$ hg clone ssh://user@dummy/repo client2 -q
$ cd client2
$ hg pull -B scratch/mybranch --traceback
pulling from ssh://user@dummy/repo
searching for changes
adding changesets
adding manifests
adding file changes
added 1 changesets with 1 changes to 1 files
new changesets 20759b6926ce (1 drafts)
(run 'hg update' to get a working copy)
$ hg log -G
o changeset: 1:20759b6926ce
| bookmark: scratch/mybranch
| tag: tip
| user: test
| date: Thu Jan 01 00:00:00 1970 +0000
| summary: scratchcommit
|
@ changeset: 0:67145f466344
user: test
date: Thu Jan 01 00:00:00 1970 +0000
summary: initialcommit
$ cd ..
Push to non-scratch bookmark
$ cd client
$ hg up 0
0 files updated, 0 files merged, 1 files removed, 0 files unresolved
$ mkcommit newcommit
created new head
$ hg push -r .
pushing to ssh://user@dummy/repo
searching for changes
remote: adding changesets
remote: adding manifests
remote: adding file changes
remote: added 1 changesets with 1 changes to 1 files
$ hg log -G -T '{desc} {phase} {bookmarks}'
@ newcommit public
|
| o scratchcommit draft scratch/mybranch
|/
o initialcommit public
Push to scratch branch
$ cd ../client2
$ hg up -q scratch/mybranch
$ mkcommit 'new scratch commit'
$ hg push -r . -B scratch/mybranch
pushing to ssh://user@dummy/repo
searching for changes
remote: pushing 2 commits:
remote: 20759b6926ce scratchcommit
remote: 1de1d7d92f89 new scratch commit
$ hg log -G -T '{desc} {phase} {bookmarks}'
@ new scratch commit draft scratch/mybranch
|
o scratchcommit draft
|
o initialcommit public
$ scratchnodes
1de1d7d92f8965260391d0513fe8a8d5973d3042 bed63daed3beba97fff2e819a148cf415c217a85
20759b6926ce827d5a8c73eb1fa9726d6f7defb2 bed63daed3beba97fff2e819a148cf415c217a85
$ scratchbookmarks
scratch/mybranch 1de1d7d92f8965260391d0513fe8a8d5973d3042
Push scratch bookmark with no new revs
$ hg push -r . -B scratch/anotherbranch
pushing to ssh://user@dummy/repo
searching for changes
remote: pushing 2 commits:
remote: 20759b6926ce scratchcommit
remote: 1de1d7d92f89 new scratch commit
$ hg log -G -T '{desc} {phase} {bookmarks}'
@ new scratch commit draft scratch/anotherbranch scratch/mybranch
|
o scratchcommit draft
|
o initialcommit public
$ scratchbookmarks
scratch/anotherbranch 1de1d7d92f8965260391d0513fe8a8d5973d3042
scratch/mybranch 1de1d7d92f8965260391d0513fe8a8d5973d3042
Pull scratch and non-scratch bookmark at the same time
$ hg -R ../repo book newbook
$ cd ../client
$ hg pull -B newbook -B scratch/mybranch --traceback
pulling from ssh://user@dummy/repo
searching for changes
adding changesets
adding manifests
adding file changes
adding remote bookmark newbook
added 1 changesets with 1 changes to 2 files
new changesets 1de1d7d92f89 (1 drafts)
(run 'hg update' to get a working copy)
$ hg log -G -T '{desc} {phase} {bookmarks}'
o new scratch commit draft scratch/mybranch
|
| @ newcommit public
| |
o | scratchcommit draft
|/
o initialcommit public
Push scratch revision without bookmark with --bundle-store
$ hg up -q tip
$ mkcommit scratchcommitnobook
$ hg log -G -T '{desc} {phase} {bookmarks}'
@ scratchcommitnobook draft
|
o new scratch commit draft scratch/mybranch
|
| o newcommit public
| |
o | scratchcommit draft
|/
o initialcommit public
$ hg push -r . --bundle-store
pushing to ssh://user@dummy/repo
searching for changes
remote: pushing 3 commits:
remote: 20759b6926ce scratchcommit
remote: 1de1d7d92f89 new scratch commit
remote: 2b5d271c7e0d scratchcommitnobook
$ hg -R ../repo log -G -T '{desc} {phase}'
o newcommit public
|
o initialcommit public
$ scratchnodes
1de1d7d92f8965260391d0513fe8a8d5973d3042 66fa08ff107451320512817bed42b7f467a1bec3
20759b6926ce827d5a8c73eb1fa9726d6f7defb2 66fa08ff107451320512817bed42b7f467a1bec3
2b5d271c7e0d25d811359a314d413ebcc75c9524 66fa08ff107451320512817bed42b7f467a1bec3
Test with pushrebase
$ mkcommit scratchcommitwithpushrebase
$ hg push -r . -B scratch/mybranch
pushing to ssh://user@dummy/repo
searching for changes
remote: pushing 4 commits:
remote: 20759b6926ce scratchcommit
remote: 1de1d7d92f89 new scratch commit
remote: 2b5d271c7e0d scratchcommitnobook
remote: d8c4f54ab678 scratchcommitwithpushrebase
$ hg -R ../repo log -G -T '{desc} {phase}'
o newcommit public
|
o initialcommit public
$ scratchnodes
1de1d7d92f8965260391d0513fe8a8d5973d3042 e3cb2ac50f9e1e6a5ead3217fc21236c84af4397
20759b6926ce827d5a8c73eb1fa9726d6f7defb2 e3cb2ac50f9e1e6a5ead3217fc21236c84af4397
2b5d271c7e0d25d811359a314d413ebcc75c9524 e3cb2ac50f9e1e6a5ead3217fc21236c84af4397
d8c4f54ab678fd67cb90bb3f272a2dc6513a59a7 e3cb2ac50f9e1e6a5ead3217fc21236c84af4397
Change the order of pushrebase and infinitepush
$ mkcommit scratchcommitwithpushrebase2
$ hg push -r . -B scratch/mybranch
pushing to ssh://user@dummy/repo
searching for changes
remote: pushing 5 commits:
remote: 20759b6926ce scratchcommit
remote: 1de1d7d92f89 new scratch commit
remote: 2b5d271c7e0d scratchcommitnobook
remote: d8c4f54ab678 scratchcommitwithpushrebase
remote: 6c10d49fe927 scratchcommitwithpushrebase2
$ hg -R ../repo log -G -T '{desc} {phase}'
o newcommit public
|
o initialcommit public
$ scratchnodes
1de1d7d92f8965260391d0513fe8a8d5973d3042 cd0586065eaf8b483698518f5fc32531e36fd8e0
20759b6926ce827d5a8c73eb1fa9726d6f7defb2 cd0586065eaf8b483698518f5fc32531e36fd8e0
2b5d271c7e0d25d811359a314d413ebcc75c9524 cd0586065eaf8b483698518f5fc32531e36fd8e0
6c10d49fe92751666c40263f96721b918170d3da cd0586065eaf8b483698518f5fc32531e36fd8e0
d8c4f54ab678fd67cb90bb3f272a2dc6513a59a7 cd0586065eaf8b483698518f5fc32531e36fd8e0
Non-fastforward scratch bookmark push
$ hg log -GT "{rev}:{node} {desc}\n"
@ 6:6c10d49fe92751666c40263f96721b918170d3da scratchcommitwithpushrebase2
|
o 5:d8c4f54ab678fd67cb90bb3f272a2dc6513a59a7 scratchcommitwithpushrebase
|
o 4:2b5d271c7e0d25d811359a314d413ebcc75c9524 scratchcommitnobook
|
o 3:1de1d7d92f8965260391d0513fe8a8d5973d3042 new scratch commit
|
| o 2:91894e11e8255bf41aa5434b7b98e8b2aa2786eb newcommit
| |
o | 1:20759b6926ce827d5a8c73eb1fa9726d6f7defb2 scratchcommit
|/
o 0:67145f4663446a9580364f70034fea6e21293b6f initialcommit
$ hg up 6c10d49fe927
0 files updated, 0 files merged, 0 files removed, 0 files unresolved
$ echo 1 > amend
$ hg add amend
$ hg ci --amend -m 'scratch amended commit'
saved backup bundle to $TESTTMP/client/.hg/strip-backup/6c10d49fe927-c99ffec5-amend.hg
$ hg log -G -T '{desc} {phase} {bookmarks}'
@ scratch amended commit draft scratch/mybranch
|
o scratchcommitwithpushrebase draft
|
o scratchcommitnobook draft
|
o new scratch commit draft
|
| o newcommit public
| |
o | scratchcommit draft
|/
o initialcommit public
$ scratchbookmarks
scratch/anotherbranch 1de1d7d92f8965260391d0513fe8a8d5973d3042
scratch/mybranch 6c10d49fe92751666c40263f96721b918170d3da
$ hg push -r . -B scratch/mybranch
pushing to ssh://user@dummy/repo
searching for changes
remote: pushing 5 commits:
remote: 20759b6926ce scratchcommit
remote: 1de1d7d92f89 new scratch commit
remote: 2b5d271c7e0d scratchcommitnobook
remote: d8c4f54ab678 scratchcommitwithpushrebase
remote: 8872775dd97a scratch amended commit
$ scratchbookmarks
scratch/anotherbranch 1de1d7d92f8965260391d0513fe8a8d5973d3042
scratch/mybranch 8872775dd97a750e1533dc1fbbca665644b32547
$ hg log -G -T '{desc} {phase} {bookmarks}'
@ scratch amended commit draft scratch/mybranch
|
o scratchcommitwithpushrebase draft
|
o scratchcommitnobook draft
|
o new scratch commit draft
|
| o newcommit public
| |
o | scratchcommit draft
|/
o initialcommit public
Check that push path is not ignored. Add new path to the hgrc
$ cat >> .hg/hgrc << EOF
> [paths]
> peer=ssh://user@dummy/client2
> EOF
Checkout last non-scrath commit
$ hg up 91894e11e8255
1 files updated, 0 files merged, 6 files removed, 0 files unresolved
$ mkcommit peercommit
Use --force because this push creates new head
$ hg push peer -r . -f
pushing to ssh://user@dummy/client2
searching for changes
remote: adding changesets
remote: adding manifests
remote: adding file changes
remote: added 2 changesets with 2 changes to 2 files (+1 heads)
$ hg -R ../repo log -G -T '{desc} {phase} {bookmarks}'
o newcommit public
|
o initialcommit public
$ hg -R ../client2 log -G -T '{desc} {phase} {bookmarks}'
o peercommit public
|
o newcommit public
|
| @ new scratch commit draft scratch/anotherbranch scratch/mybranch
| |
| o scratchcommit draft
|/
o initialcommit public