Bug #379: old heartbeat peers not marked down - Ceph - Ceph

Actions

Copy link

Bug #379

closed

old heartbeat peers not marked down

Added by Sage Weil over 13 years ago. Updated over 13 years ago.

Status:

Resolved

Priority:

Normal

Assignee:

Category:

OSD

Target version:

v0.21.2

% Done:

Source:

Tags:

Backport:

Regression:

Severity:

Reviewed:

Affected Versions:

ceph-qa-suite:

Pull request ID:

Crash signature (v1):

Crash signature (v2):

Description

On Wido's osd1 I see lots of

10.08.26_06:08:12.923443 7f95e8300710 -- [2001:16f8:10:2::c3c3:a24f]:6801/1030 >> [2001:16f8:10:2::c3c3:3f6c]:6801/8617 pipe(0x4639500 sd=38 pgs=5230 cs=2 l=0).connect claims to be [2001:16f8:10:2::c3c3:3f6c]:6801/12092 not [2001:16f8:10:2::c3c3:3f6c]:6801/8617 - wrong node!

That's the prior incarnation of osd6's heartbeat address:

root@logger:~# ceph osd dump -o - | grep 2001:16f8:10:2::c3c3:3f6c
osd6 in weight 1 up   (up_from 10374 up_thru 10820 down_at 10373 last_clean 10307-10372) [2001:16f8:10:2::c3c3:3f6c]:6800/12092 [2001:16f8:10:2::c3c3:3f6c]:6801/12092
root@logger:~# ceph osd dump -o - 10372| grep 2001:16f8:10:2::c3c3:3f6c
osd6 in weight 1 up   (up_from 10307 up_thru 10370 down_at 10306 last_clean 10208-10305) [2001:16f8:10:2::c3c3:3f6c]:6800/8617 [2001:16f8:10:2::c3c3:3f6c]:6801/8617