Project

General

Profile

Tasks #921

Bug #910: Multi-MDS Ceph does not pass fsstress

Snaprealm issue

Added by Greg Farnum about 9 years ago. Updated about 9 years ago.

Status:
Resolved
Priority:
Normal
Assignee:
Category:
-
Target version:
% Done:

0%

Tags:
Reviewed:
Affected Versions:
Pull request ID:

Description

mds/MDCache.cc: In function 'void MDCache::handle_dentry_unlink(MDentryUnlink*)', in thread '0x7f182c0eb710'
mds/MDCache.cc: 8760: FAILED assert(straydn->first >= in->first)
 ceph version 0.25-386-g15ab9e7 (commit:15ab9e7396ad55249c6fe8a7ffd3c2f77199a5dd)
 1: (MDCache::handle_dentry_unlink(MDentryUnlink*)+0x6dd) [0x58185d]
 2: (MDS::_dispatch(Message*)+0x20c4) [0x4b96e4]
 3: (MDS::ms_dispatch(Message*)+0x5b) [0x4ba3fb]
 4: (SimpleMessenger::dispatch_entry()+0x89a) [0x48955a]
 5: (SimpleMessenger::DispatchThread::entry()+0x1c) [0x48222c]
 6: (()+0x68ba) [0x7f182e7408ba]
 7: (clone()+0x6d) [0x7f182d3d502d]

Sage thinks (if I got this right) that this might just be the dentry or inode getting initialized improperly before sending since it's a stray (or maybe on receipt?).

History

#1 Updated by Greg Farnum about 9 years ago

  • Subject changed from Failed assert in handle_dentry_unlink to Snaprealm issue

For some reason in a long chain of passing the inode back and forth and stuff it somehow gets its own snaprealm with a seq of 0. Logs and core dump are in kai:~gregf/logs/fsstress/snaprealm_bug. We'll look at it later.

#2 Updated by Greg Farnum about 9 years ago

  • Status changed from New to In Progress

Sage and I looked at this and think he got it in a series of patches. Will update with testing.

#3 Updated by Greg Farnum about 9 years ago

  • Status changed from In Progress to 7

#4 Updated by Greg Farnum about 9 years ago

  • Status changed from 7 to Resolved

Also available in: Atom PDF