Bug #3612
disconnect_session_watches assert(pg) failed
% Done:
0%
Source:
Development
Tags:
Backport:
Regression:
No
Severity:
3 - minor
Reviewed:
Affected Versions:
ceph-qa-suite:
Pull request ID:
Crash signature:
Description
osd/OSD.cc: 2434: FAILED assert(pg) 1: (ceph::__ceph_assert_fail(char const*, char const*, int, char const*)+0x95) [0x11a4ac1] 2: (OSD::disconnect_session_watches(OSD::Session*)+0x2a7) [0xea6ef5] 3: (OSD::ms_handle_reset(Connection*)+0x155) [0xea761d] 4: (Messenger::ms_deliver_handle_reset(Connection*)+0x4b) [0x126b419] 5: (DispatchQueue::entry()+0x176) [0x126a4be] 6: (DispatchQueue::DispatchThread::entry()+0x1c) [0x118ac14] 7: (Thread::_entry_func(void*)+0x23) [0x11932ad] 8: (()+0x7e9a) [0x7fc73f45de9a] 9: (clone()+0x6d) [0x7fc73d5e84bd]
On teuthology, in /a/sage-2012-12-11_19:47:13-rados-wip-3459-testing-basic/11975 (osd.4, plana86). Sam has started looking at it already.
Related issues
History
#1 Updated by Samuel Just about 8 years ago
The assert is bad, no guarantee that the pg is still around after we drop the watch_lock. remove_watchers_and_notifies would have been called in the intervening journey from Primary-Active to Stray-Purged.