Bug #13422: pid file that was deleted after ceph restart leads to osd process lose control - Ceph - Ceph

Actions

Copy link

Bug #13422

closed

pid file that was deleted after ceph restart leads to osd process lose control

Added by shun song over 8 years ago. Updated over 7 years ago.

Status:

Resolved

Priority:

High

Assignee:

Loïc Dachary

Category:

Target version:

% Done:

Source:

other

Tags:

Backport:

hammer,infernalis

Regression:

Severity:

3 - minor

Reviewed:

Affected Versions:

v.91

ceph-qa-suite:

Pull request ID:

Crash signature (v1):

Crash signature (v2):

Description

when restarting ceph, the same osd will start twice. the first time is regular start;the other is activated by ceph-disk activated all. if the second startup starts fast enough before the pid file created. Unluckly, only one process can get fsid_lock and live through. then the lose process will delete the pid file when exiting. As a result, this osd may stop unregularlly when using 'service ceph stop osd.{id}'.

resovled method:
locking pid file using file lock

Related issues 4 (0 open — 4 closed)

Actions

Copy link

Updated by huang jun over 8 years ago

we also met this many times, even if https://github.com/ceph/ceph/commit/487c22a8a4b3ba099f9c19125c720e99e7c8d0db was patched

Actions

Copy link

Updated by shun song over 8 years ago

how about not to remove pid-file,when process exits, becuase the pid-file is important for juding whether osd/mon is running. at the same time, to avoid compelling to write the pid-file when multiple times start up one osd or monitor, lock is recommanded.

Actions

Copy link

Updated by Nathan Cutler over 8 years ago

master PR: https://github.com/ceph/ceph/pull/6763

Actions

Copy link

Updated by Nathan Cutler over 8 years ago

Status changed from New to Fix Under Review

Actions

Copy link

Updated by Kefu Chai over 8 years ago

Has duplicate Bug #13238: duplicate start of ceph-osd daemon added

Actions

Copy link

Updated by Kefu Chai about 8 years ago

https://github.com/ceph/ceph/pull/7075

Actions

Copy link

Updated by Loïc Dachary about 8 years ago

Backport set to hammer,infernalis

Actions

Copy link

Updated by Shinobu Kinjo about 8 years ago

Will patch be (able to be) backported to 0.80.8?

Actions

Copy link

Updated by Kefu Chai about 8 years ago

Will patch be (able to be) backported to 0.80.8?

@Shinobu 0.80.8 is a firefly release, and since firefly is retired, see http://docs.ceph.com/docs/master/releases/ , we are not going to backport to this release.

and yes, it is able to be backported to 0.80.8, but this will not happen in the upstream.

Actions

Copy link

#10