Project

General

Profile

Actions
Copy link

Bug #9895

closed

Master/giant branch: OSD deadlock during recovery

Added by Andrey Korolyov over 9 years ago. Updated over 9 years ago.

Status:
Duplicate
Priority:
High
Assignee:
-
Category:
OSD
Target version:
-
% Done:

0%

Source:
Community (user)
Tags:
Backport:
Regression:
Severity:
3 - minor
Reviewed:
Affected Versions:
ceph-qa-suite:
Pull request ID:
Crash signature (v1):
Crash signature (v2):

Description

Given eight-OSD, two-node cluster (node01 and node04), three mons (node01, node04, twin2). OSDs placed on node04 acts as a data senders, backfilling kv-based OSDs on node01. Sometimes one of those senders reaches deadlock (strace and thread dump attached), leaving cluster effectively (and making it stale due reaching zero online copies for some objects).

ceph.conf.gz can be found there: http://lists.ceph.com/pipermail/ceph-users-ceph.com/attachments/20141022/33c6c485/attachment-0003.bin

ceph version 0.86-104-gb05efdd (b05efddb77290b86eb5c150776c761ab84f66f37)

Files

Download all files
ceph-osd-9-stale.threaddump.txt (57.2 KB) ceph-osd-9-stale.threaddump.txt backtrace Andrey Korolyov, 10/26/2014 11:24 AM
ceph-state-map-osd9-stale.txt (6.95 KB) ceph-state-map-osd9-stale.txt ceph -s output Andrey Korolyov, 10/26/2014 11:24 AM
osd9-out.strace.txt (62 Bytes) osd9-out.strace.txt strace output Andrey Korolyov, 10/26/2014 11:24 AM

Related issues 1 (0 open — 1 closed)

Is duplicate of Messengers - Bug #9898: osd: fast dispatch deadlock in mark_down (giant)Resolved10/26/2014

Actions
Actions
Copy link
 #1

Updated by Sage Weil over 9 years ago

  • Status changed from New to Duplicate
Actions
Copy link

Also available in: Atom PDF