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 repo
$ cd repo
$ i=0; while [ "$i" -lt 213 ]; do echo a >> a; i=`expr $i + 1`; done
$ hg add a
$ cp a b
$ hg add b
Wide diffstat:
$ hg diff --stat
a | 213 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
b | 213 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
2 files changed, 426 insertions(+), 0 deletions(-)
diffstat width:
$ COLUMNS=24 hg diff --config ui.interactive=true --stat
a | 213 ++++++++++++++
b | 213 ++++++++++++++
2 files changed, 426 insertions(+), 0 deletions(-)
$ hg ci -m adda
$ cat >> a <<EOF
> a
> a
> a
> EOF
Narrow diffstat:
$ hg diff --stat
a | 3 +++
1 files changed, 3 insertions(+), 0 deletions(-)
$ hg ci -m appenda
>>> open("c", "wb").write(b"\0") and None
$ touch d
$ hg add c d
Binary diffstat:
$ hg diff --stat
c | Bin
1 files changed, 0 insertions(+), 0 deletions(-)
Binary git diffstat:
$ hg diff --stat --git
c | Bin
d | 0
2 files changed, 0 insertions(+), 0 deletions(-)
$ hg ci -m createb
>>> open("file with spaces", "wb").write(b"\0") and None
$ hg add "file with spaces"
Filename with spaces diffstat:
$ hg diff --stat
file with spaces | Bin
1 files changed, 0 insertions(+), 0 deletions(-)
Filename with spaces git diffstat:
$ hg diff --stat --git
file with spaces | Bin
1 files changed, 0 insertions(+), 0 deletions(-)
Filename without "a/" or "b/" (issue5759):
$ hg diff --config 'diff.noprefix=1' -c1 --stat --git
a | 3 +++
1 files changed, 3 insertions(+), 0 deletions(-)
$ hg diff --config 'diff.noprefix=1' -c2 --stat --git
c | Bin
d | 0
2 files changed, 0 insertions(+), 0 deletions(-)
$ hg log --config 'diff.noprefix=1' -r '1:' -p --stat --git
changeset: 1:3a95b07bb77f
user: test
date: Thu Jan 01 00:00:00 1970 +0000
summary: appenda
a | 3 +++
1 files changed, 3 insertions(+), 0 deletions(-)
diff --git a a
--- a
+++ a
@@ -211,3 +211,6 @@
a
a
a
+a
+a
+a
changeset: 2:c60a6c753773
tag: tip
user: test
date: Thu Jan 01 00:00:00 1970 +0000
summary: createb
c | Bin
d | 0
2 files changed, 0 insertions(+), 0 deletions(-)
diff --git c c
new file mode 100644
index e69de29bb2d1d6434b8b29ae775ad8c2e48c5391..f76dd238ade08917e6712764a16a22005a50573d
GIT binary patch
literal 1
Ic${MZ000310RR91
diff --git d d
new file mode 100644
diffstat within directories:
$ hg rm -f 'file with spaces'
$ mkdir dir1 dir2
$ echo new1 > dir1/new
$ echo new2 > dir2/new
$ hg add dir1/new dir2/new
$ hg diff --stat
dir1/new | 1 +
dir2/new | 1 +
2 files changed, 2 insertions(+), 0 deletions(-)
$ hg diff --stat --root dir1
new | 1 +
1 files changed, 1 insertions(+), 0 deletions(-)
$ hg diff --stat --root dir1 dir2
warning: dir2 not inside relative root dir1
$ hg diff --stat --root dir1 -I dir1/old
$ cd dir1
$ hg diff --stat .
dir1/new | 1 +
1 files changed, 1 insertions(+), 0 deletions(-)
$ hg diff --stat . --config ui.relative-paths=yes
new | 1 +
1 files changed, 1 insertions(+), 0 deletions(-)
$ hg diff --stat --root .
new | 1 +
1 files changed, 1 insertions(+), 0 deletions(-)
$ hg diff --stat --root . --config ui.relative-paths=yes
new | 1 +
1 files changed, 1 insertions(+), 0 deletions(-)
--root trumps ui.relative-paths
$ hg diff --stat --root .. --config ui.relative-paths=yes
new | 1 +
../dir2/new | 1 +
2 files changed, 2 insertions(+), 0 deletions(-)
$ hg diff --stat --root ../dir1 ../dir2
warning: ../dir2 not inside relative root .
$ hg diff --stat --root . -I old
$ cd ..
Files with lines beginning with '--' or '++' should be properly counted in diffstat
$ hg up -Cr tip
0 files updated, 0 files merged, 0 files removed, 0 files unresolved
$ rm dir1/new
$ rm dir2/new
$ rm "file with spaces"
$ cat > file << EOF
> line 1
> line 2
> line 3
> EOF
$ hg commit -Am file
adding file
Lines added starting with '--' should count as additions
$ cat > file << EOF
> line 1
> -- line 2, with dashes
> line 3
> EOF
$ hg diff --root .
diff -r be1569354b24 file
--- a/file Thu Jan 01 00:00:00 1970 +0000
+++ b/file * (glob)
@@ -1,3 +1,3 @@
line 1
-line 2
+-- line 2, with dashes
line 3
$ hg diff --root . --stat
file | 2 +-
1 files changed, 1 insertions(+), 1 deletions(-)
Lines changed starting with '--' should count as deletions
$ hg commit -m filev2
$ cat > file << EOF
> line 1
> -- line 2, with dashes, changed again
> line 3
> EOF
$ hg diff --root .
diff -r 160f7c034df6 file
--- a/file Thu Jan 01 00:00:00 1970 +0000
+++ b/file * (glob)
@@ -1,3 +1,3 @@
line 1
--- line 2, with dashes
+-- line 2, with dashes, changed again
line 3
$ hg diff --root . --stat
file | 2 +-
1 files changed, 1 insertions(+), 1 deletions(-)
Lines changed starting with '--' should count as deletions
and starting with '++' should count as additions
$ cat > file << EOF
> line 1
> ++ line 2, switched dashes to plusses
> line 3
> EOF
$ hg diff --root .
diff -r 160f7c034df6 file
--- a/file Thu Jan 01 00:00:00 1970 +0000
+++ b/file * (glob)
@@ -1,3 +1,3 @@
line 1
--- line 2, with dashes
+++ line 2, switched dashes to plusses
line 3
$ hg diff --root . --stat
file | 2 +-
1 files changed, 1 insertions(+), 1 deletions(-)
When a file is renamed, --git shouldn't loss the info about old file
$ hg init issue6025
$ cd issue6025
$ echo > a
$ hg ci -Am 'add a'
adding a
$ hg mv a b
$ hg diff --git
diff --git a/a b/b
rename from a
rename to b
$ hg diff --stat
a | 1 -
b | 1 +
2 files changed, 1 insertions(+), 1 deletions(-)
$ hg diff --stat --git
a => b | 0
1 files changed, 0 insertions(+), 0 deletions(-)
-- filename may contain whitespaces
$ echo > c
$ hg ci -Am 'add c'
adding c
$ hg mv c 'new c'
$ hg diff --git
diff --git a/c b/new c
rename from c
rename to new c
$ hg diff --stat
c | 1 -
new c | 1 +
2 files changed, 1 insertions(+), 1 deletions(-)
$ hg diff --stat --git
c => new c | 0
1 files changed, 0 insertions(+), 0 deletions(-)
Make sure `diff --stat -q --config diff.git-0` shows stat (issue4037)
$ hg status
A new c
R c
$ hg diff --stat -q
c | 1 -
new c | 1 +
2 files changed, 1 insertions(+), 1 deletions(-)