Bug #2128
filestore: check() fails during sync
% Done:
0%
Source:
Development
Tags:
Backport:
Regression:
No
Severity:
3 - minor
Reviewed:
Affected Versions:
ceph-qa-suite:
Pull request ID:
Crash signature (v1):
Crash signature (v2):
Description
2012-03-02 15:17:02.277344 7f1fed615700 filestore(/tmp/cephtest/data/osd.0.data) Missing: seq 431 *** Caught signal (Aborted) ** in thread 7f1fed615700 ceph version 0.42.2-226-g9dae7ed (commit:9dae7ed77ab7179d46db63b4bdd577b99ad7b742) 1: (ceph::BackTrace::BackTrace(int)+0x2d) [0xae24fb] 2: /tmp/cephtest/binary/usr/local/bin/ceph-osd() [0xc2cac8] 3: (()+0xfb40) [0x7f1ff529eb40] 4: (gsignal()+0x35) [0x7f1ff386eba5] 5: (abort()+0x180) [0x7f1ff38726b0] 6: (__assert_fail()+0xf1) [0x7f1ff3867a71] 7: (FileStore::sync_entry()+0xd6e) [0xd26c56] 8: (FileStore::SyncThread::entry()+0x1c) [0xd348d4] 9: (Thread::_entry_func(void*)+0x23) [0xad4dc1] 10: (()+0x7971) [0x7f1ff5296971]
workload is thrashing, so there are a bunch of
2012-03-02 15:07:01.356510 7f1feb611700 filestore(/tmp/cephtest/data/osd.0.data) _omap_setkeys temp/3270c60b/mds0_sessionmap/head
calls, but no actual client data.
incidentally, we should probably avoid creating keys for objects that don't have omap values.. it looks like this might be populating leveldb with intermediate nodes or something?
History
#1 Updated by Sage Weil about 12 years ago
could it be 75cbed61e94a7974e40230360c6781d85f47576d ?
#2 Updated by Sage Weil about 12 years ago
- Status changed from New to Rejected
actually, i hit this running the new code with the filestore transaction leak bug fix reverted (to reproduce the btrfs issue). i haven't seen this on master, so I'm guessing it's related to that.