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
$ cat <<EOF >> $HGRCPATH
> [extensions]
> purge =
> EOF
init
$ hg init t
$ cd t
setup
$ echo r1 > r1
$ hg ci -qAmr1 -d'0 0'
$ mkdir directory
$ echo r2 > directory/r2
$ hg ci -qAmr2 -d'1 0'
$ echo 'ignored' > .hgignore
$ hg ci -qAmr3 -d'2 0'
delete an empty directory
$ mkdir empty_dir
$ hg purge -p -v
empty_dir
$ hg purge -v
removing directory empty_dir
$ ls
directory
r1
delete an untracked directory
$ mkdir untracked_dir
$ touch untracked_dir/untracked_file1
$ touch untracked_dir/untracked_file2
$ hg purge -p
untracked_dir/untracked_file1
untracked_dir/untracked_file2
$ hg purge -v
removing file untracked_dir/untracked_file1
removing file untracked_dir/untracked_file2
removing directory untracked_dir
$ ls
directory
r1
delete an untracked file
$ touch untracked_file
$ touch untracked_file_readonly
$ "$PYTHON" <<EOF
> import os
> import stat
> f = 'untracked_file_readonly'
> os.chmod(f, stat.S_IMODE(os.stat(f).st_mode) & ~stat.S_IWRITE)
> EOF
$ hg purge -p
untracked_file
untracked_file_readonly
$ hg purge -v
removing file untracked_file
removing file untracked_file_readonly
$ ls
directory
r1
delete an untracked file in a tracked directory
$ touch directory/untracked_file
$ hg purge -p
directory/untracked_file
$ hg purge -v
removing file directory/untracked_file
$ ls
directory
r1
delete nested directories
$ mkdir -p untracked_directory/nested_directory
$ hg purge -p
untracked_directory/nested_directory
$ hg purge -v
removing directory untracked_directory/nested_directory
removing directory untracked_directory
$ ls
directory
r1
delete nested directories from a subdir
$ mkdir -p untracked_directory/nested_directory
$ cd directory
$ hg purge -p
untracked_directory/nested_directory
$ hg purge -v
removing directory untracked_directory/nested_directory
removing directory untracked_directory
$ cd ..
$ ls
directory
r1
delete only part of the tree
$ mkdir -p untracked_directory/nested_directory
$ touch directory/untracked_file
$ cd directory
$ hg purge -p ../untracked_directory
untracked_directory/nested_directory
$ hg purge -v ../untracked_directory
removing directory untracked_directory/nested_directory
removing directory untracked_directory
$ cd ..
$ ls
directory
r1
$ ls directory/untracked_file
directory/untracked_file
$ rm directory/untracked_file
skip ignored files if --all not specified
$ touch ignored
$ hg purge -p
$ hg purge -v
$ ls
directory
ignored
r1
$ hg purge -p --all
ignored
$ hg purge -v --all
removing file ignored
$ ls
directory
r1
abort with missing files until we support name mangling filesystems
$ touch untracked_file
$ rm r1
hide error messages to avoid changing the output when the text changes
$ hg purge -p 2> /dev/null
untracked_file
$ hg st
! r1
? untracked_file
$ hg purge -p
untracked_file
$ hg purge -v 2> /dev/null
removing file untracked_file
$ hg st
! r1
$ hg purge -v
$ hg revert --all --quiet
$ hg st -a
tracked file in ignored directory (issue621)
$ echo directory >> .hgignore
$ hg ci -m 'ignore directory'
$ touch untracked_file
$ hg purge -p
untracked_file
$ hg purge -v
removing file untracked_file
skip excluded files
$ touch excluded_file
$ hg purge -p -X excluded_file
$ hg purge -v -X excluded_file
$ ls
directory
excluded_file
r1
$ rm excluded_file
skip files in excluded dirs
$ mkdir excluded_dir
$ touch excluded_dir/file
$ hg purge -p -X excluded_dir
$ hg purge -v -X excluded_dir
$ ls
directory
excluded_dir
r1
$ ls excluded_dir
file
$ rm -R excluded_dir
skip excluded empty dirs
$ mkdir excluded_dir
$ hg purge -p -X excluded_dir
$ hg purge -v -X excluded_dir
$ ls
directory
excluded_dir
r1
$ rmdir excluded_dir
skip patterns
$ mkdir .svn
$ touch .svn/foo
$ mkdir directory/.svn
$ touch directory/.svn/foo
$ hg purge -p -X .svn -X '*/.svn'
$ hg purge -p -X re:.*.svn
$ rm -R .svn directory r1
only remove files
$ mkdir -p empty_dir dir
$ touch untracked_file dir/untracked_file
$ hg purge -p --files
dir/untracked_file
untracked_file
$ hg purge -v --files
removing file dir/untracked_file
removing file untracked_file
$ ls
dir
empty_dir
$ ls dir
only remove dirs
$ mkdir -p empty_dir dir
$ touch untracked_file dir/untracked_file
$ hg purge -p --dirs
empty_dir
$ hg purge -v --dirs
removing directory empty_dir
$ ls
dir
untracked_file
$ ls dir
untracked_file
remove both files and dirs
$ mkdir -p empty_dir dir
$ touch untracked_file dir/untracked_file
$ hg purge -p --files --dirs
dir/untracked_file
untracked_file
empty_dir
$ hg purge -v --files --dirs
removing file dir/untracked_file
removing file untracked_file
removing directory empty_dir
removing directory dir
$ ls
$ cd ..