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
from __future__ import absolute_import
import glob
import os
import shutil
import stat
import tempfile
import unittest
from mercurial import (
pycompat,
util,
)
atomictempfile = util.atomictempfile
if pycompat.ispy3:
xrange = range
class testatomictempfile(unittest.TestCase):
def setUp(self):
self._testdir = tempfile.mkdtemp(b'atomictempfiletest')
self._filename = os.path.join(self._testdir, b'testfilename')
def tearDown(self):
shutil.rmtree(self._testdir, True)
def testsimple(self):
file = atomictempfile(self._filename)
self.assertFalse(os.path.isfile(self._filename))
tempfilename = file._tempname
self.assertTrue(
tempfilename
in glob.glob(os.path.join(self._testdir, b'.testfilename-*'))
)
file.write(b'argh\n')
file.close()
self.assertTrue(os.path.isfile(self._filename))
self.assertTrue(
tempfilename
not in glob.glob(os.path.join(self._testdir, b'.testfilename-*'))
)
# discard() removes the temp file without making the write permanent
def testdiscard(self):
file = atomictempfile(self._filename)
(dir, basename) = os.path.split(file._tempname)
file.write(b'yo\n')
file.discard()
self.assertFalse(os.path.isfile(self._filename))
self.assertTrue(basename not in os.listdir(b'.'))
# if a programmer screws up and passes bad args to atomictempfile, they
# get a plain ordinary TypeError, not infinite recursion
def testoops(self):
with self.assertRaises(TypeError):
atomictempfile()
# checkambig=True avoids ambiguity of timestamp
def testcheckambig(self):
def atomicwrite(checkambig):
f = atomictempfile(self._filename, checkambig=checkambig)
f.write(b'FOO')
f.close()
# try some times, because reproduction of ambiguity depends on
# "filesystem time"
for i in xrange(5):
atomicwrite(False)
oldstat = os.stat(self._filename)
if oldstat[stat.ST_CTIME] != oldstat[stat.ST_MTIME]:
# subsequent changing never causes ambiguity
continue
repetition = 3
# repeat atomic write with checkambig=True, to examine
# whether st_mtime is advanced multiple times as expected
for j in xrange(repetition):
atomicwrite(True)
newstat = os.stat(self._filename)
if oldstat[stat.ST_CTIME] != newstat[stat.ST_CTIME]:
# timestamp ambiguity was naturally avoided while repetition
continue
# st_mtime should be advanced "repetition" times, because
# all atomicwrite() occurred at same time (in sec)
oldtime = (oldstat[stat.ST_MTIME] + repetition) & 0x7FFFFFFF
self.assertTrue(newstat[stat.ST_MTIME] == oldtime)
# no more examination is needed, if assumption above is true
break
else:
# This platform seems too slow to examine anti-ambiguity
# of file timestamp (or test happened to be executed at
# bad timing). Exit silently in this case, because running
# on other faster platforms can detect problems
pass
def testread(self):
with open(self._filename, 'wb') as f:
f.write(b'foobar\n')
file = atomictempfile(self._filename, mode=b'rb')
self.assertTrue(file.read(), b'foobar\n')
file.discard()
def testcontextmanagersuccess(self):
"""When the context closes, the file is closed"""
with atomictempfile(b'foo') as f:
self.assertFalse(os.path.isfile(b'foo'))
f.write(b'argh\n')
self.assertTrue(os.path.isfile(b'foo'))
def testcontextmanagerfailure(self):
"""On exception, the file is discarded"""
try:
with atomictempfile(b'foo') as f:
self.assertFalse(os.path.isfile(b'foo'))
f.write(b'argh\n')
raise ValueError
except ValueError:
pass
self.assertFalse(os.path.isfile(b'foo'))
if __name__ == '__main__':
import silenttestrunner
silenttestrunner.main(__name__)