Actions
Bug #10743
closedmds/MDLog.cc: 259: FAILED assert(!capped) on MDS rank shutdown
% Done:
0%
Source:
other
Tags:
Backport:
Regression:
Severity:
3 - minor
Reviewed:
Affected Versions:
ceph-qa-suite:
Component(FS):
Labels (FS):
Pull request ID:
Crash signature (v1):
Crash signature (v2):
Description
Hacked MDCache::eval-stray to never purge if my rank >0
Vstart cluster with 2 MDSs
mkdir ALPHA
dmesg> ALPHA/test
ceph daemon mds.a export dir /ALPHA 1
rm -rf ALPHA
ceph mds set_max_mds 1
ceph mds deactivate 1
boom (log attached)
This is probably intermittent.
Files
Updated by Greg Farnum about 9 years ago
Probably means there's some point where we run shutdown_pass() and assume it succeeds. (That's what calls MDLog::cap().)
Updated by John Spray about 9 years ago
The trouble seems to be that rank 1 is sending lock messages to revoke rank 0's replica of ~mds1, then proceeding to cap, then the acks for the lock messages come in after the cap. We need to make the shutdown sequence wait for locking activity on mydir to quiesce before capping the log
Updated by John Spray about 9 years ago
- Status changed from New to In Progress
- Assignee set to John Spray
Updated by John Spray about 9 years ago
- Status changed from In Progress to Fix Under Review
Updated by Greg Farnum about 9 years ago
- Status changed from Fix Under Review to Resolved
Actions