Project

General

Profile

Actions
Copy link

Bug #2395

closed

kernel crash after unmap a rdb device while the cluster is down

Added by Danny Kukawka almost 12 years ago. Updated almost 12 years ago.

Status:
Resolved
Priority:
High
Assignee:
-
Category:
libceph
Target version:
-
% Done:

0%

Source:
Development
Tags:
Backport:
Regression:
Severity:
Reviewed:
Affected Versions:
ceph-qa-suite:
Crash signature (v1):
Crash signature (v2):

Description

1) get a ceph cluster running
2) create a RBD and map it to an client
3) shutdown the cluster
4) as soon as the cluster is down: unmap the RBD on the cluster
-> kernel crash

Notes:
  • this happens with kernel v3.0 directly after the unmap call returns
  • it also happens with the code from git branch remotes/origin/for-linus but the behavior is different
    -> as soon as you call unmap on the rbd, the command will never come back
    and you see libceph looping over the MONs for ever. As soon as you try
    to stop/kill the rbd call , you will end up in a kernel crash
  • ceph version was v0.45 @ SLES11-SP2

Files

crash-kernel-v3.3.0 (6.22 KB) crash-kernel-v3.3.0 Danny Kukawka, 05/11/2012 11:05 AM
Actions
Copy link
 #1

Updated by Sage Weil almost 12 years ago

I am unable to reproduce this with the current code. There was a lot of rbd setup/teardown code that got cleaned up for v3.4. Is it possible to test again that kernel?

Actions
Copy link
 #2

Updated by Sage Weil almost 12 years ago

  • Status changed from New to Resolved

I'm going go assume this is running the older code and close it. If not, let us know!

Actions
Copy link

Also available in: Atom PDF