Project

General

Profile

Bug #718

sync hangs

Added by Sage Weil almost 9 years ago. Updated almost 9 years ago.

Status:
Resolved
Priority:
Urgent
Assignee:
Category:
Clustered MDS
Target version:
% Done:

0%

Spent time:
Source:
Tags:
Backport:
Regression:
No
Severity:
3 - minor
Reviewed:
Affected Versions:
ceph-qa-suite:
Crash signature:

Description

There were a few reports, and I'm also seeing it on current unstable. Haven't checked other branches yet.

In one case, reproducing was as simple as echo foo > mnt/foo ; sync, although later that did work. This time it was a partial kernel untar, control-z, sync.


Related issues

Duplicated by Linux kernel client - Bug #227: caps flush races with caps export Resolved 06/24/2010

History

#1 Updated by Sage Weil almost 9 years ago

on v2.6.37,


[  802.320000] libceph:  osd_client.c:1421 : sync done (thru tid 7008)
[  802.320000] ceph:     mds_client.c:3098 : sync
[  802.320000] ceph:     mds_client.c:3103 : sync want tid 16370 flush_seq 8705
[  802.320000] ceph:           caps.c:2886 : flush_dirty_caps
[  802.320000] ceph:     mds_client.c:3055 : wait_unsafe_requests want 16370
[  802.320000] ceph:     mds_client.c:3088 : wait_unsafe_requests done
[  802.320000] ceph:     mds_client.c:1249 : check_cap_flush want 8705
[  802.320000] ceph:     mds_client.c:310  : mdsc get_session 000000008076a7f0 1 -> 2
[  802.320000] ceph:     mds_client.c:1272 : check_cap_flush still flushing 000000007b0d3d40 seq 2415 <= 8705 to mds0
[  802.320000] ceph:     mds_client.c:321  : mdsc put_session 000000008076a7f0 2 -> 1
[  802.320000] ceph:     mds_client.c:1249 : check_cap_flush want 8705
[  802.320000] ceph:     mds_client.c:310  : mdsc get_session 000000008076a7f0 1 -> 2
[  802.320000] ceph:     mds_client.c:1272 : check_cap_flush still flushing 000000007b0d3d40 seq 2415 <= 8705 to mds0
... and then no more mention check_cap_flush

So the problem appears to be with the mds cap flushing... and either server-side or present is 2.6.37.

#2 Updated by Sage Weil almost 9 years ago

  • Category set to Clustered MDS
  • Status changed from New to Resolved
  • Assignee set to Sage Weil

fixed in unstable branch, commit:55ee8fe37598475ed363d078cb50d19e0524c69f 'ceph: fix flushing of caps vs cap import'

Also available in: Atom PDF