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
$ echo "[extensions]" >> $HGRCPATH
$ echo "mq=" >> $HGRCPATH
make a test repository that looks like this:
o 2:28bc7b1afd6a
|
| @ 1:d7fe2034f71b
|/
o 0/62ecad8b70e5
$ hg init r0
$ cd r0
$ touch f0
$ hg ci -m0 -Aq
$ touch f1
$ hg ci -m1 -Aq
$ hg update 0 -q
$ touch f2
$ hg ci -m2 -Aq
$ hg update 1 -q
make some patches with a parent: 1:d7fe2034f71b -> p0 -> p1
$ echo cp0 >> fp0
$ hg add fp0
$ hg ci -m p0 -d "0 0"
$ hg export -r. > p0
$ hg strip -qn .
$ hg qimport p0
adding p0 to series file
$ hg qpush
applying p0
now at: p0
$ echo cp1 >> fp1
$ hg add fp1
$ hg qnew p1 -d "0 0"
$ hg qpop -aq
patch queue now empty
qpush --exact when at the parent
$ hg update 1 -q
$ hg qpush -e
applying p0
now at: p0
$ hg parents -qr qbase
1:d7fe2034f71b
$ hg qpop -aq
patch queue now empty
$ hg qpush -e p0
applying p0
now at: p0
$ hg parents -qr qbase
1:d7fe2034f71b
$ hg qpop -aq
patch queue now empty
$ hg qpush -e p1
applying p0
applying p1
now at: p1
$ hg parents -qr qbase
1:d7fe2034f71b
$ hg qpop -aq
patch queue now empty
qpush --exact when at another rev
$ hg update 0 -q
$ hg qpush -e
1 files updated, 0 files merged, 0 files removed, 0 files unresolved
applying p0
now at: p0
$ hg parents -qr qbase
1:d7fe2034f71b
$ hg qpop -aq
patch queue now empty
$ hg update 0 -q
$ hg qpush -e p0
1 files updated, 0 files merged, 0 files removed, 0 files unresolved
applying p0
now at: p0
$ hg parents -qr qbase
1:d7fe2034f71b
$ hg qpop -aq
patch queue now empty
$ hg update 0 -q
$ hg qpush -e p1
1 files updated, 0 files merged, 0 files removed, 0 files unresolved
applying p0
applying p1
now at: p1
$ hg parents -qr qbase
1:d7fe2034f71b
$ hg qpop -aq
patch queue now empty
$ hg update 0 -q
$ hg qpush -ea
1 files updated, 0 files merged, 0 files removed, 0 files unresolved
applying p0
applying p1
now at: p1
$ hg parents -qr qbase
1:d7fe2034f71b
$ hg qpop -aq
patch queue now empty
qpush --exact while crossing branches
$ hg update 2 -q
$ hg qpush -e
1 files updated, 0 files merged, 1 files removed, 0 files unresolved
applying p0
now at: p0
$ hg parents -qr qbase
1:d7fe2034f71b
$ hg qpop -aq
patch queue now empty
$ hg update 2 -q
$ hg qpush -e p0
1 files updated, 0 files merged, 1 files removed, 0 files unresolved
applying p0
now at: p0
$ hg parents -qr qbase
1:d7fe2034f71b
$ hg qpop -aq
patch queue now empty
$ hg update 2 -q
$ hg qpush -e p1
1 files updated, 0 files merged, 1 files removed, 0 files unresolved
applying p0
applying p1
now at: p1
$ hg parents -qr qbase
1:d7fe2034f71b
$ hg qpop -aq
patch queue now empty
$ hg update 2 -q
$ hg qpush -ea
1 files updated, 0 files merged, 1 files removed, 0 files unresolved
applying p0
applying p1
now at: p1
$ hg parents -qr qbase
1:d7fe2034f71b
$ hg qpop -aq
patch queue now empty
qpush --exact --force with changes to an unpatched file
$ hg update 1 -q
$ echo c0 >> f0
$ hg qpush -e
abort: local changes found
[255]
$ hg qpush -ef
applying p0
now at: p0
$ cat f0
c0
$ rm f0
$ touch f0
$ hg qpop -aq
patch queue now empty
$ hg update 1 -q
$ echo c0 >> f0
$ hg qpush -e p1
abort: local changes found
[255]
$ hg qpush -e p1 -f
applying p0
applying p1
now at: p1
$ cat f0
c0
$ rm f0
$ touch f0
$ hg qpop -aq
patch queue now empty
qpush --exact --force with changes to a patched file
$ hg update 1 -q
$ echo cp0-bad >> fp0
$ hg add fp0
$ hg qpush -e
abort: local changes found
[255]
$ hg qpush -ef
applying p0
file fp0 already exists
1 out of 1 hunks FAILED -- saving rejects to file fp0.rej
patch failed, unable to continue (try -v)
patch failed, rejects left in working directory
errors during apply, please fix and qrefresh p0
[2]
$ cat fp0
cp0-bad
$ cat fp0.rej
--- fp0
+++ fp0
@@ -0,0 +1,1 @@
+cp0
$ hg qpop -aqf
patch queue now empty
$ rm fp0
$ rm fp0.rej
$ hg update 1 -q
$ echo cp1-bad >> fp1
$ hg add fp1
$ hg qpush -e p1
abort: local changes found
[255]
$ hg qpush -e p1 -f
applying p0
applying p1
file fp1 already exists
1 out of 1 hunks FAILED -- saving rejects to file fp1.rej
patch failed, unable to continue (try -v)
patch failed, rejects left in working directory
errors during apply, please fix and qrefresh p1
[2]
$ cat fp1
cp1-bad
$ cat fp1.rej
--- fp1
+++ fp1
@@ -0,0 +1,1 @@
+cp1
$ hg qpop -aqf
patch queue now empty
$ hg forget fp1
$ rm fp1
$ rm fp1.rej
qpush --exact when already at a patch
$ hg update 1
0 files updated, 0 files merged, 0 files removed, 0 files unresolved
$ hg qpush -e p0
applying p0
now at: p0
$ hg qpush -e p1
abort: cannot push --exact with applied patches
[255]
$ hg qpop -aq
patch queue now empty
qpush --exact --move should fail
$ hg qpush -e --move p1
abort: cannot use --exact and --move together
[255]
qpush --exact a patch without a parent recorded
$ hg qpush -q
now at: p0
$ grep -v '# Parent' .hg/patches/p0 > p0.new
$ mv p0.new .hg/patches/p0
$ hg qpop -aq
patch queue now empty
$ hg qpush -e
abort: p0 does not have a parent recorded
[255]
$ hg qpush -e p0
abort: p0 does not have a parent recorded
[255]
$ hg qpush -e p1
abort: p0 does not have a parent recorded
[255]
$ hg qpush -ea
abort: p0 does not have a parent recorded
[255]
$ cd ..