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
$ hg init a
$ cd a
$ echo a > a
$ hg ci -A -d'1 0' -m a
adding a
$ cd ..
$ hg init b
$ cd b
$ echo b > b
$ hg ci -A -d'1 0' -m b
adding b
$ cd ..
$ hg clone a c
updating to branch default
1 files updated, 0 files merged, 0 files removed, 0 files unresolved
$ cd c
$ cat >> .hg/hgrc <<EOF
> [paths]
> relative = ../a
> EOF
$ hg pull -f ../b
pulling from ../b
searching for changes
warning: repository is unrelated
requesting all changes
adding changesets
adding manifests
adding file changes
added 1 changesets with 1 changes to 1 files (+1 heads)
new changesets b6c483daf290
(run 'hg heads' to see heads, 'hg merge' to merge)
$ hg merge
1 files updated, 0 files merged, 0 files removed, 0 files unresolved
(branch merge, don't forget to commit)
$ cd ..
Testing -R/--repository:
$ hg -R a tip
changeset: 0:8580ff50825a
tag: tip
user: test
date: Thu Jan 01 00:00:01 1970 +0000
summary: a
$ hg --repository b tip
changeset: 0:b6c483daf290
tag: tip
user: test
date: Thu Jan 01 00:00:01 1970 +0000
summary: b
-R with a URL:
$ hg -R file:a identify
8580ff50825a tip
$ hg -R file://localhost/`pwd`/a/ identify
8580ff50825a tip
-R with path aliases:
$ cd c
$ hg -R default identify
8580ff50825a tip
$ hg -R relative identify
8580ff50825a tip
$ echo '[paths]' >> $HGRCPATH
$ echo 'relativetohome = a' >> $HGRCPATH
$ HOME=`pwd`/../ hg -R relativetohome identify
8580ff50825a tip
$ cd ..
#if no-outer-repo
Implicit -R:
$ hg ann a/a
0: a
$ hg ann a/a a/a
0: a
$ hg ann a/a b/b
abort: no repository found in '$TESTTMP' (.hg not found)!
[255]
$ hg -R b ann a/a
abort: a/a not under root '$TESTTMP/b'
(consider using '--cwd b')
[255]
$ hg log
abort: no repository found in '$TESTTMP' (.hg not found)!
[255]
#endif
Abbreviation of long option:
$ hg --repo c tip
changeset: 1:b6c483daf290
tag: tip
parent: -1:000000000000
user: test
date: Thu Jan 01 00:00:01 1970 +0000
summary: b
earlygetopt with duplicate options (36d23de02da1):
$ hg --cwd a --cwd b --cwd c tip
changeset: 1:b6c483daf290
tag: tip
parent: -1:000000000000
user: test
date: Thu Jan 01 00:00:01 1970 +0000
summary: b
$ hg --repo c --repository b -R a tip
changeset: 0:8580ff50825a
tag: tip
user: test
date: Thu Jan 01 00:00:01 1970 +0000
summary: a
earlygetopt short option without following space:
$ hg -q -Rb tip
0:b6c483daf290
earlygetopt with illegal abbreviations:
$ hg --confi "foo.bar=baz"
abort: option --config may not be abbreviated!
[255]
$ hg --cw a tip
abort: option --cwd may not be abbreviated!
[255]
$ hg --rep a tip
abort: option -R has to be separated from other options (e.g. not -qR) and --repository may only be abbreviated as --repo!
[255]
$ hg --repositor a tip
abort: option -R has to be separated from other options (e.g. not -qR) and --repository may only be abbreviated as --repo!
[255]
$ hg -qR a tip
abort: option -R has to be separated from other options (e.g. not -qR) and --repository may only be abbreviated as --repo!
[255]
$ hg -qRa tip
abort: option -R has to be separated from other options (e.g. not -qR) and --repository may only be abbreviated as --repo!
[255]
Testing --cwd:
$ hg --cwd a parents
changeset: 0:8580ff50825a
tag: tip
user: test
date: Thu Jan 01 00:00:01 1970 +0000
summary: a
Testing -y/--noninteractive - just be sure it is parsed:
$ hg --cwd a tip -q --noninteractive
0:8580ff50825a
$ hg --cwd a tip -q -y
0:8580ff50825a
Testing -q/--quiet:
$ hg -R a -q tip
0:8580ff50825a
$ hg -R b -q tip
0:b6c483daf290
$ hg -R c --quiet parents
0:8580ff50825a
1:b6c483daf290
Testing -v/--verbose:
$ hg --cwd c head -v
changeset: 1:b6c483daf290
tag: tip
parent: -1:000000000000
user: test
date: Thu Jan 01 00:00:01 1970 +0000
files: b
description:
b
changeset: 0:8580ff50825a
user: test
date: Thu Jan 01 00:00:01 1970 +0000
files: a
description:
a
$ hg --cwd b tip --verbose
changeset: 0:b6c483daf290
tag: tip
user: test
date: Thu Jan 01 00:00:01 1970 +0000
files: b
description:
b
Testing --config:
$ hg --cwd c --config paths.quuxfoo=bar paths | grep quuxfoo > /dev/null && echo quuxfoo
quuxfoo
$ hg --cwd c --config '' tip -q
abort: malformed --config option: '' (use --config section.name=value)
[255]
$ hg --cwd c --config a.b tip -q
abort: malformed --config option: 'a.b' (use --config section.name=value)
[255]
$ hg --cwd c --config a tip -q
abort: malformed --config option: 'a' (use --config section.name=value)
[255]
$ hg --cwd c --config a.= tip -q
abort: malformed --config option: 'a.=' (use --config section.name=value)
[255]
$ hg --cwd c --config .b= tip -q
abort: malformed --config option: '.b=' (use --config section.name=value)
[255]
Testing --debug:
$ hg --cwd c log --debug
changeset: 1:b6c483daf2907ce5825c0bb50f5716226281cc1a
tag: tip
phase: public
parent: -1:0000000000000000000000000000000000000000
parent: -1:0000000000000000000000000000000000000000
manifest: 1:23226e7a252cacdc2d99e4fbdc3653441056de49
user: test
date: Thu Jan 01 00:00:01 1970 +0000
files+: b
extra: branch=default
description:
b
changeset: 0:8580ff50825a50c8f716709acdf8de0deddcd6ab
phase: public
parent: -1:0000000000000000000000000000000000000000
parent: -1:0000000000000000000000000000000000000000
manifest: 0:a0c8bcbbb45c63b90b70ad007bf38961f64f2af0
user: test
date: Thu Jan 01 00:00:01 1970 +0000
files+: a
extra: branch=default
description:
a
Testing --traceback:
#if no-chg
$ hg --cwd c --config x --traceback id 2>&1 | grep -i 'traceback'
Traceback (most recent call last):
Traceback (most recent call last): (py3 !)
#else
Traceback for '--config' errors not supported with chg.
$ hg --cwd c --config x --traceback id 2>&1 | grep -i 'traceback'
[1]
#endif
Testing --time:
$ hg --cwd a --time id
8580ff50825a tip
time: real * (glob)
Testing --version:
$ hg --version -q
Mercurial Distributed SCM * (glob)
hide outer repo
$ hg init
Testing -h/--help:
#if no-extraextensions
$ hg -h
Mercurial Distributed SCM
list of commands:
Repository creation:
clone make a copy of an existing repository
init create a new repository in the given directory
Remote repository management:
incoming show new changesets found in source
outgoing show changesets not found in the destination
paths show aliases for remote repositories
pull pull changes from the specified source
push push changes to the specified destination
serve start stand-alone webserver
Change creation:
commit commit the specified files or all outstanding changes
Change manipulation:
backout reverse effect of earlier changeset
graft copy changes from other branches onto the current branch
merge merge another revision into working directory
Change organization:
bookmarks create a new bookmark or list existing bookmarks
branch set or show the current branch name
branches list repository named branches
phase set or show the current phase name
tag add one or more tags for the current or given revision
tags list repository tags
File content management:
annotate show changeset information by line for each file
cat output the current or given revision of files
copy mark files as copied for the next commit
diff diff repository (or selected files)
grep search for a pattern in specified files
Change navigation:
bisect subdivision search of changesets
heads show branch heads
identify identify the working directory or specified revision
log show revision history of entire repository or files
Working directory management:
add add the specified files on the next commit
addremove add all new files, delete all missing files
files list tracked files
forget forget the specified files on the next commit
remove remove the specified files on the next commit
rename rename files; equivalent of copy + remove
resolve redo merges or set/view the merge status of files
revert restore files to their checkout state
root print the root (top) of the current working directory
shelve save and set aside changes from the working directory
status show changed files in the working directory
summary summarize working directory state
unshelve restore a shelved change to the working directory
update update working directory (or switch revisions)
Change import/export:
archive create an unversioned archive of a repository revision
bundle create a bundle file
export dump the header and diffs for one or more changesets
import import an ordered set of patches
unbundle apply one or more bundle files
Repository maintenance:
manifest output the current or given revision of the project manifest
recover roll back an interrupted transaction
verify verify the integrity of the repository
Help:
config show combined config settings from all hgrc files
help show help for a given topic or a help overview
version output version and copyright information
additional help topics:
Mercurial identifiers:
filesets Specifying File Sets
hgignore Syntax for Mercurial Ignore Files
patterns File Name Patterns
revisions Specifying Revisions
urls URL Paths
Mercurial output:
color Colorizing Outputs
dates Date Formats
diffs Diff Formats
templating Template Usage
Mercurial configuration:
config Configuration Files
environment Environment Variables
extensions Using Additional Features
flags Command-line flags
hgweb Configuring hgweb
merge-tools Merge Tools
pager Pager Support
Concepts:
bundlespec Bundle File Formats
glossary Glossary
phases Working with Phases
subrepos Subrepositories
Miscellaneous:
deprecated Deprecated Features
internals Technical implementation topics
scripting Using Mercurial from scripts and automation
(use 'hg help -v' to show built-in aliases and global options)
$ hg --help
Mercurial Distributed SCM
list of commands:
Repository creation:
clone make a copy of an existing repository
init create a new repository in the given directory
Remote repository management:
incoming show new changesets found in source
outgoing show changesets not found in the destination
paths show aliases for remote repositories
pull pull changes from the specified source
push push changes to the specified destination
serve start stand-alone webserver
Change creation:
commit commit the specified files or all outstanding changes
Change manipulation:
backout reverse effect of earlier changeset
graft copy changes from other branches onto the current branch
merge merge another revision into working directory
Change organization:
bookmarks create a new bookmark or list existing bookmarks
branch set or show the current branch name
branches list repository named branches
phase set or show the current phase name
tag add one or more tags for the current or given revision
tags list repository tags
File content management:
annotate show changeset information by line for each file
cat output the current or given revision of files
copy mark files as copied for the next commit
diff diff repository (or selected files)
grep search for a pattern in specified files
Change navigation:
bisect subdivision search of changesets
heads show branch heads
identify identify the working directory or specified revision
log show revision history of entire repository or files
Working directory management:
add add the specified files on the next commit
addremove add all new files, delete all missing files
files list tracked files
forget forget the specified files on the next commit
remove remove the specified files on the next commit
rename rename files; equivalent of copy + remove
resolve redo merges or set/view the merge status of files
revert restore files to their checkout state
root print the root (top) of the current working directory
shelve save and set aside changes from the working directory
status show changed files in the working directory
summary summarize working directory state
unshelve restore a shelved change to the working directory
update update working directory (or switch revisions)
Change import/export:
archive create an unversioned archive of a repository revision
bundle create a bundle file
export dump the header and diffs for one or more changesets
import import an ordered set of patches
unbundle apply one or more bundle files
Repository maintenance:
manifest output the current or given revision of the project manifest
recover roll back an interrupted transaction
verify verify the integrity of the repository
Help:
config show combined config settings from all hgrc files
help show help for a given topic or a help overview
version output version and copyright information
additional help topics:
Mercurial identifiers:
filesets Specifying File Sets
hgignore Syntax for Mercurial Ignore Files
patterns File Name Patterns
revisions Specifying Revisions
urls URL Paths
Mercurial output:
color Colorizing Outputs
dates Date Formats
diffs Diff Formats
templating Template Usage
Mercurial configuration:
config Configuration Files
environment Environment Variables
extensions Using Additional Features
flags Command-line flags
hgweb Configuring hgweb
merge-tools Merge Tools
pager Pager Support
Concepts:
bundlespec Bundle File Formats
glossary Glossary
phases Working with Phases
subrepos Subrepositories
Miscellaneous:
deprecated Deprecated Features
internals Technical implementation topics
scripting Using Mercurial from scripts and automation
(use 'hg help -v' to show built-in aliases and global options)
#endif
Not tested: --debugger