mds: just wait the client flushes the snap and dirty buffer by lxbsz · Pull Request #53238 · ceph/ceph

lxbsz · 2023-08-31T09:40:45Z

When truncating the inode we will just set the ifile lock state to LOCK_XLOCKSNAP and then try to revoke the 'Fb' caps, but if the client couldn't release the 'Fb' cap in time just replies with a normal cap updating request, the MDS will successfully transfer the ifile's lock state to LOCK_EXCL, which is stable.

That means the MDS will wake up the truncating request and continue truncating the objects from Rados without waiting the clients to flush the diry buffer.

Fixes: commit 9c65920 ("mds: force client flush snap data before
truncating objects")
Fixes: https://tracker.ceph.com/issues/62580

Contribution Guidelines

To sign and title your commits, please refer to Submitting Patches to Ceph.
If you are submitting a fix for a stable branch (e.g. "pacific"), please refer to Submitting Patches to Ceph - Backports for the proper workflow.

Checklist

Tracker (select at least one)
- References tracker ticket
- Very recent bug; references commit where it was introduced
- New feature (ticket optional)
- Doc update (no ticket needed)
- Code cleanup (no ticket needed)
Component impact
- Affects Dashboard, opened tracker ticket
- Affects Orchestrator, opened tracker ticket
- No impact that needs to be tracked
Documentation (select at least one)
- Updates relevant documentation
- No doc update is appropriate
Tests (select at least one)
- Includes unit test(s)
- Includes integration test(s)
- Includes bug reproducer
- No tests

Show available Jenkins commands

jenkins retest this please
jenkins test classic perf
jenkins test crimson perf
jenkins test signed
jenkins test make check
jenkins test make check arm64
jenkins test submodules
jenkins test dashboard
jenkins test dashboard cephadm
jenkins test api
jenkins test docs
jenkins render docs
jenkins test ceph-volume all
jenkins test ceph-volume tox
jenkins test windows

vshankar · 2023-09-19T07:58:28Z

jenkins test windows

vshankar

Fix makes sense.

vshankar · 2023-10-11T04:20:58Z

jenkins test windows

vshankar

https://tracker.ceph.com/projects/cephfs/wiki/Main#13-Oct-2023

vshankar · 2023-10-20T05:12:37Z

jenkins test windows

vshankar · 2023-10-20T09:58:18Z

jenkins test windows

vshankar · 2023-10-23T02:08:23Z

@lxbsz could you please rebase and push - jenkins test are failing to pass :/

lxbsz · 2023-10-23T02:49:36Z

@lxbsz could you please rebase and push - jenkins test are failing to pass :/

Done.

lxbsz · 2023-10-23T02:52:25Z

jenkins test make check arm64

vshankar · 2023-11-02T09:53:16Z

jenkins test windows

vshankar · 2023-11-06T13:49:18Z

@lxbsz please rebase and push - jenkins mess :/

lxbsz · 2023-11-07T01:14:17Z

@lxbsz please rebase and push - jenkins mess :/

Done.

lxbsz · 2023-11-07T01:17:46Z

jenkins test make check arm64

vshankar · 2023-11-07T04:47:20Z

Windows failure looks real

[2023-11-07T03:02:11.000Z] [isolated][googletest] ceph_test_libcephfs failed. Error: Command timed out (1800s): "cmd /c 'C:\ceph\ceph_test_libcephfs.exe --gtest_output=xml:C:\workspace\test_results\out\ceph_test_libcephfs\ceph_test_libcephfs_results.xml --gtest_filter="-LibCephFS.Deleg*" >> C:\workspace\test_results\out\ceph_test_libcephfs\ceph_test_libcephfs_results.log 2>&1'".

@petrutlucian94 @lxbsz

petrutlucian94 · 2023-11-07T09:25:21Z

Windows failure looks real

[2023-11-07T03:02:11.000Z] [isolated][googletest] ceph_test_libcephfs failed. Error: Command timed out (1800s): "cmd /c 'C:\ceph\ceph_test_libcephfs.exe --gtest_output=xml:C:\workspace\test_results\out\ceph_test_libcephfs\ceph_test_libcephfs_results.xml --gtest_filter="-LibCephFS.Deleg*" >> C:\workspace\test_results\out\ceph_test_libcephfs\ceph_test_libcephfs_results.log 2>&1'".

@petrutlucian94 @lxbsz

LibCephFS.SnapDiffLib2 seems to hang, I'll try running it locally.

petrutlucian94 · 2023-11-07T13:43:03Z

It's not a Windows issue, it hangs on Linux as well. It's just that make check doesn't run ceph_test_libcephfs. Here's a trace:

Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
bt
__futex_abstimed_wait_common64 (private=0, cancel=true, abstime=0x0, op=393, expected=0, futex_word=0x7fff89826648) at ./nptl/futex-internal.c:57
57	./nptl/futex-internal.c: No such file or directory.
(gdb) bt
#0  __futex_abstimed_wait_common64 (private=0, cancel=true, abstime=0x0, op=393, expected=0, 
    futex_word=0x7fff89826648) at ./nptl/futex-internal.c:57
#1  __futex_abstimed_wait_common (cancel=true, private=0, abstime=0x0, clockid=0, expected=0, 
    futex_word=0x7fff89826648) at ./nptl/futex-internal.c:87
#2  __GI___futex_abstimed_wait_cancelable64 (futex_word=futex_word@entry=0x7fff89826648, 
    expected=expected@entry=0, clockid=clockid@entry=0, abstime=abstime@entry=0x0, 
    private=private@entry=0) at ./nptl/futex-internal.c:139
#3  0x00007f5739384a41 in __pthread_cond_wait_common (abstime=0x0, clockid=0, mutex=0x561a234fe950, 
    cond=0x7fff89826620) at ./nptl/pthread_cond_wait.c:503
#4  ___pthread_cond_wait (cond=0x7fff89826620, mutex=0x561a234fe950) at ./nptl/pthread_cond_wait.c:627
#5  0x00007f573a4f7e7d in std::condition_variable::wait<Client::make_request(MetaRequest*, const UserPerm&, InodeRef*, bool*, mds_rank_t, ceph::bufferlist*, size_t)::<lambda()> > (__p=..., __lock=..., 
    this=0x7fff89826620) at /usr/include/c++/11/condition_variable:103
#6  Client::make_request (this=<optimized out>, request=0x561a235c72b0, perms=..., ptarget=0x0, 
    pcreated=0x0, use_mds=<optimized out>, pdirbl=0x0, feature_needed=18446744073709551615)
    at /home/ubuntu/workspace/ceph.linux/src/client/Client.cc:2053
#7  0x00007f573a514fdc in Client::_do_setattr (this=0x561a234fe7d0, in=0x7f5714013310, 
    stx=<optimized out>, mask=32, perms=..., inp=0x0, aux=0x0)
    at /home/ubuntu/workspace/ceph.linux/src/client/Client.cc:8308
#8  0x00007f573a51937c in Client::__setattrx (this=0x561a234fe7d0, in=0x7f5714013310, 
    stx=0x7fff89826950, mask=32, perms=..., inp=<optimized out>)
    at /home/ubuntu/workspace/ceph.linux/src/client/Client.cc:8343
#9  0x00007f573a53e091 in Client::_setattrx (this=<optimized out>, in=..., stx=<optimized out>, 
    mask=<optimized out>, perms=...) at /home/ubuntu/workspace/ceph.linux/src/client/Client.cc:8363
#10 0x00007f573a53fa05 in Client::setattrx (this=0x561a234fe7d0, relpath=<optimized out>, 
    stx=0x7fff89826950, mask=32, perms=..., flags=0)
    at /home/ubuntu/workspace/ceph.linux/src/client/Client.cc:8423
#11 0x00007f573a53fc16 in Client::truncate (this=<optimized out>, relpath=<optimized out>, 
    length=<optimized out>, perms=...) at /home/ubuntu/workspace/ceph.linux/src/client/Client.cc:11720
#12 0x0000561a2144ae55 in TestMount::write_full (this=0x7fff89826bc0, relpath=<optimized out>, 
    data="hello world to be removed at snap3") at /usr/include/c++/11/bits/basic_string.h:194
#13 0x0000561a21437daf in TestMount::prepareSnapDiffLib2Cases (this=0x7fff89826bc0)
    at /usr/include/c++/11/ext/new_allocator.h:79
#14 0x0000561a21444c67 in LibCephFS_SnapDiffLib2_Test::TestBody (this=<optimized out>)
    at /home/ubuntu/workspace/ceph.linux/src/test/libcephfs/snapdiff.cc:817
#15 0x0000561a21480fef in testing::internal::HandleSehExceptionsInMethodIfSupported<testing::Test, void>
    (location=0x561a2149483a "the test body", method=<optimized out>, object=0x561a235c22a0)
    at /home/ubuntu/workspace/ceph.linux/src/googletest/googletest/src/gtest.cc:2605
#16 testing::internal::HandleExceptionsInMethodIfSupported<testing::Test, void> (
    object=object@entry=0x561a235c22a0, method=<optimized out>, 
    location=location@entry=0x561a2149483a "the test body")
    at /home/ubuntu/workspace/ceph.linux/src/googletest/googletest/src/gtest.cc:2641
#17 0x0000561a214756f6 in testing::Test::Run (this=0x561a235c22a0)
    at /home/ubuntu/workspace/ceph.linux/src/googletest/googletest/src/gtest.cc:2680
#18 testing::Test::Run (this=0x561a235c22a0)
    at /home/ubuntu/workspace/ceph.linux/src/googletest/googletest/src/gtest.cc:2670
#19 0x0000561a21475855 in testing::TestInfo::Run (this=0x561a23387830)
    at /home/ubuntu/workspace/ceph.linux/src/googletest/googletest/src/gtest.cc:2858
#20 testing::TestInfo::Run (this=0x561a23387830)
    at /home/ubuntu/workspace/ceph.linux/src/googletest/googletest/src/gtest.cc:2831

lxbsz · 2023-11-08T00:08:57Z

It's not a Windows issue, it hangs on Linux as well. It's just that make check doesn't run ceph_test_libcephfs. Here's a trace:
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
bt

[...]

#20 testing::TestInfo::Run (this=0x561a23387830)
at /home/ubuntu/workspace/ceph.linux/src/googletest/googletest/src/gtest.cc:2831

@petrutlucian94 BTW, did the above test include this PR ? Or just the
upstream ?

lxbsz · 2023-11-08T00:11:43Z

jenkins test make check arm64

petrutlucian94 · 2023-11-08T07:50:33Z

The above test fails if I apply this commit but passes if I revert it, so there's something wrong with the commit.

./ceph_test_libcephfs --gtest_filter="*SnapDiffLib2*"
Note: Google Test filter = *SnapDiffLib2*
[==========] Running 1 test from 1 test suite.
[----------] Global test environment set-up.
[----------] 1 test from LibCephFS
[ RUN      ] LibCephFS.SnapDiffLib2
2 vs. 3 vs. 4
---------snap1 listing verification---------
...
------------- closing -------------
[       OK ] LibCephFS.SnapDiffLib2 (11042 ms)
[----------] 1 test from LibCephFS (11042 ms total)

[----------] Global test environment tear-down
[==========] 1 test from 1 test suite ran. (11042 ms total)
[  PASSED  ] 1 test.

$ git log -5 --oneline
3a1ad72220e (HEAD) Revert "mds: just wait the client flushes the snap and dirty buffer"
034512a6e5c (lxbsz/wip-62580) mds: just wait the client flushes the snap and dirty buffer
f76333be1a2 Merge pull request #54197 from ivancich/wip-bucket-stats-add-gen
125cc9406ab Merge pull request #54365 from cbodley/wip-63455
c6b83286033 Merge PR #53568 into main

lxbsz · 2023-11-08T09:35:04Z

The above test fails if I apply this commit but passes if I revert it, so there's something wrong with the commit.

./ceph_test_libcephfs --gtest_filter="*SnapDiffLib2*"
Note: Google Test filter = *SnapDiffLib2*
[==========] Running 1 test from 1 test suite.
[----------] Global test environment set-up.
[----------] 1 test from LibCephFS
[ RUN      ] LibCephFS.SnapDiffLib2
2 vs. 3 vs. 4
---------snap1 listing verification---------
...
------------- closing -------------
[       OK ] LibCephFS.SnapDiffLib2 (11042 ms)
[----------] 1 test from LibCephFS (11042 ms total)

[----------] Global test environment tear-down
[==========] 1 test from 1 test suite ran. (11042 ms total)
[  PASSED  ] 1 test.

$ git log -5 --oneline
3a1ad72220e (HEAD) Revert "mds: just wait the client flushes the snap and dirty buffer"
034512a6e5c (lxbsz/wip-62580) mds: just wait the client flushes the snap and dirty buffer
f76333be1a2 Merge pull request #54197 from ivancich/wip-bucket-stats-add-gen
125cc9406ab Merge pull request #54365 from cbodley/wip-63455
c6b83286033 Merge PR #53568 into main

Okay, I will have a look later. Thanks @petrutlucian94

When truncating the inode we will just set the ifile lock state to LOCK_XLOCKSNAP and then try to revoke the 'Fb' caps, but if the client couldn't release the 'Fb' cap in time just replies with a normal cap updating request, the MDS will successfully transfer the ifile's lock state to LOCK_EXCL, which is stable. That means the MDS will wake up the truncating request and continue truncating the objects from Rados without waiting the clients to flush the diry buffer. Fixes: commit 9c65920 ("mds: force client flush snap data before truncating objects") Fixes: https://tracker.ceph.com/issues/62580 Signed-off-by: Xiubo Li <xiubli@redhat.com>

lxbsz · 2023-11-15T08:02:46Z

Fixed it. @vshankar Please review it again, thanks!

vshankar · 2023-11-22T05:37:47Z

Fixed it. @vshankar Please review it again, thanks!

I've picked it up for fs suite runs @lxbsz. What was the cause of the failure BTW?

lxbsz · 2023-11-22T06:49:46Z

Fixed it. @vshankar Please review it again, thanks!

I've picked it up for fs suite runs @lxbsz. What was the cause of the failure BTW?

The old code couldn't correctly handle the corner cases and couldn't wake up the waiters. The new fixes just make it more precise and only do this just in case in LOCK_XLOCKSNAP state.

vshankar · 2023-11-27T10:28:27Z

Test runs in ~2h - wip-vshankar-testing-20231127.102654

vshankar · 2023-12-04T09:29:08Z

https://pulpito.ceph.com/?branch=wip-vshankar-testing-20231127.102654

(rhel pkg install failures are a bunch, so, this would need a revalidate)

vshankar · 2024-01-12T05:44:55Z

https://pulpito.ceph.com/?branch=wip-vshankar-testing-20240103.072409-1

vshankar

https://tracker.ceph.com/projects/cephfs/wiki/Main#17th-Jan-2024

lxbsz requested review from a team, batrick and vshankar August 31, 2023 09:40

github-actions bot added the cephfs Ceph File System label Aug 31, 2023

lxbsz mentioned this pull request Aug 31, 2023

mds: drop locks and retry when lock set changes #52522

Merged

14 tasks

vshankar reviewed Oct 11, 2023

View reviewed changes

vshankar added the wip-vshankar-testing4 label Oct 11, 2023

vshankar approved these changes Oct 20, 2023

View reviewed changes

lxbsz force-pushed the wip-62580 branch from 2a19545 to 0af28f4 Compare October 23, 2023 02:49

vshankar added ready-to-merge and removed wip-vshankar-testing4 labels Nov 2, 2023

lxbsz force-pushed the wip-62580 branch from 0af28f4 to 034512a Compare November 7, 2023 01:13

lxbsz force-pushed the wip-62580 branch from 034512a to 68179ae Compare November 15, 2023 07:29

vshankar added wip-vshankar-testing and removed ready-to-merge labels Nov 22, 2023

vshankar approved these changes Jan 17, 2024

View reviewed changes

vshankar removed the wip-vshankar-testing label Jan 18, 2024

vshankar merged commit 2c51672 into ceph:main Jan 18, 2024

This was referenced Feb 26, 2024

reef: mds: just wait the client flushes the snap and dirty buffer #55743

Merged

quincy: mds: just wait the client flushes the snap and dirty buffer #55744

Merged

Conversation

lxbsz commented Aug 31, 2023

Contribution Guidelines

Checklist

Uh oh!

vshankar commented Sep 19, 2023

Uh oh!

vshankar left a comment

Choose a reason for hiding this comment

Uh oh!

vshankar commented Oct 11, 2023

Uh oh!

vshankar left a comment

Choose a reason for hiding this comment

Uh oh!

vshankar commented Oct 20, 2023

Uh oh!

vshankar commented Oct 20, 2023

Uh oh!

vshankar commented Oct 23, 2023

Uh oh!

lxbsz commented Oct 23, 2023

Uh oh!

lxbsz commented Oct 23, 2023

Uh oh!

vshankar commented Nov 2, 2023

Uh oh!

vshankar commented Nov 6, 2023

Uh oh!

lxbsz commented Nov 7, 2023

Uh oh!

lxbsz commented Nov 7, 2023

Uh oh!

vshankar commented Nov 7, 2023

Uh oh!

petrutlucian94 commented Nov 7, 2023

Uh oh!

petrutlucian94 commented Nov 7, 2023

Uh oh!

lxbsz commented Nov 8, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

lxbsz commented Nov 8, 2023

Uh oh!

petrutlucian94 commented Nov 8, 2023

Uh oh!

lxbsz commented Nov 8, 2023

Uh oh!

lxbsz commented Nov 15, 2023

Uh oh!

vshankar commented Nov 22, 2023

Uh oh!

lxbsz commented Nov 22, 2023

Uh oh!

vshankar commented Nov 27, 2023

Uh oh!

vshankar commented Dec 4, 2023

Uh oh!

vshankar commented Jan 12, 2024

Uh oh!

vshankar left a comment

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

lxbsz commented Nov 8, 2023 •

edited

Loading