Project

General

Profile

Actions
Copy link

Bug #38718

open

'osd crush weight-set create-compat' (and other OSDMonitor commands) can leak uncommitted state

Added by Sage Weil about 5 years ago. Updated over 4 years ago.

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

0%

Source:
Tags:
Backport:
Regression:
No
Severity:
3 - minor
Reviewed:
Affected Versions:
ceph-qa-suite:
Component(RADOS):
Pull request ID:
Crash signature (v1):
Crash signature (v2):

Description

2019-03-13T04:04:25.173 INFO:teuthology.orchestra.run.smithi204:> sudo adjust-ulimits ceph-coverage /home/ubuntu/cephtest/archive/coverage timeout 120 ceph --cluster ceph quorum_status
2019-03-13T04:04:25.292 INFO:tasks.workunit.client.0.smithi022.stderr:compat weight-set already created
2019-03-13T04:04:25.305 INFO:tasks.workunit.client.0.smithi022.stderr:+ ceph osd crush weight-set ls
2019-03-13T04:04:25.306 INFO:tasks.workunit.client.0.smithi022.stderr:+ grep '(compat)'

can fail. /a/sage-2019-03-13_02:19:41-rados-wip-sage3-testing-2019-03-12-1657-distro-basic-smithi/3715050

like a zillion other places in OSDMonitor.cc, we look at the pending crush map and possibly reply with no action. We need to repeat the checks on the committed osd/crush map.

Actions
Copy link
 #1

Updated by Patrick Donnelly over 4 years ago

  • Status changed from 12 to New
Actions
Copy link

Also available in: Atom PDF