Project

General

Profile

Bug #15432

kcephfs: umount -f can fail after mds reconnect failure

Added by Sage Weil almost 4 years ago. Updated about 2 months ago.

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

0%

Source:
other
Tags:
Backport:
Regression:
No
Severity:
3 - minor
Reviewed:
Affected Versions:
ceph-qa-suite:
Crash signature:

Description

We've seen this several times on the lab cluster and teuthology:

- fail to reconnect to mds for some reason
- umount f
mount doesn't go away..

History

#2 Updated by Zheng Yan almost 4 years ago

find a bug that may cause forced umount hang
https://github.com/ceph/ceph-client/commit/e9344de458cd61efac6cccb98c785164b53d9c61

In my test, cleanup dirty data/metadata and mds requests work well. But it still can't cleanup pages under writeback. 'umount -f' hangs if there are hang OSD requests,

#3 Updated by Ian Colle over 3 years ago

  • Assignee set to Zheng Yan

Zheng - is this still an issue?

#4 Updated by Zheng Yan over 3 years ago

  • Status changed from New to 12
  • Priority changed from Urgent to Normal

It's not completed resolved. 'umount -f' hangs if there are hang OSD requests (caused by unhealth OSD cluster or libceph bug)

#5 Updated by Zheng Yan almost 3 years ago

base on Jeff's ENOSPC work, It should be easy to implement function that abort pending osd requests for 'umount -f'

#6 Updated by Patrick Donnelly about 2 months ago

  • Status changed from 12 to New

Also available in: Atom PDF