Bug #328: MDS crash: MDCache::remove_inode(CInode*) - CephFS - Ceph

Actions

Copy link

Bug #328

closed

MDS crash: MDCache::remove_inode(CInode*)

Added by Wido den Hollander over 13 years ago. Updated over 7 years ago.

Status:

Resolved

Priority:

Normal

Assignee:

Category:

Target version:

% Done:

Source:

Tags:

Backport:

Regression:

Severity:

Reviewed:

Affected Versions:

ceph-qa-suite:

Component(FS):

Labels (FS):

Pull request ID:

Crash signature (v1):

Crash signature (v2):

Description

Today i tried to sync kernel.org again, this went fine, until my log parition filled up and mds0 got stuck.

Somehow the kernel client did not switch to mds1, so this got stuck.

I killed mds0, cleaned up the logs and tried to start it again, this then failed:

mds/MDCache.cc: In function 'void MDCache::remove_inode(CInode*)':
mds/MDCache.cc:230: FAILED assert(o->get_num_ref() == 0)
 1: (EMetaBlob::replay(MDS*, LogSegment*)+0x100d) [0x625d6d]
 2: (EUpdate::replay(MDS*)+0x1f) [0x62b67f]
 3: (MDLog::_replay_thread()+0x700) [0x61a3d0]
 4: (MDLog::ReplayThread::entry()+0xd) [0x4a54bd]
 5: (Thread::_entry_func(void*)+0xa) [0x4883ba]
 6: (()+0x69ca) [0x7f9d8ffe29ca]
 7: (clone()+0x6d) [0x7f9d8f2016fd]
 NOTE: a copy of the executable, or `objdump -rdS <executable>` is needed to interpret this.

Tried starting it several times, but kept crashing with the same errors.

The core, binary and logs are available on logger.ceph.widodh.nl in /srv/ceph/issues/cmds_crash_mdcache_remove_inode