Project

General

Profile

Actions

Bug #40112

closed

mon: rados/multimon tests fail with clock skew

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

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

0%

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

Description

See

http://pulpito.ceph.com/sage-2019-05-30_21:14:09-rados:multimon-master-distro-basic-smithi/

or

http://pulpito.ceph.com/sage-2019-06-03_14:22:18-rados:multimon-master-distro-basic-smithi/
http://pulpito.ceph.com/sage-2019-06-03_14:24:56-rados:multimon-nautilus-distro-basic-smithi/

The second variant, where we don't see the SKEW failure, appears to be caused because the ceph setup command to set the crush tunables takes foreeever due to no quorum, but eventually the clocks correct themselves and then the job proceeds (probably NTP)?

The underlying problem appears to be that mons can't form quorum if there is too much skew, which prevents a meaningful test.

whenever mon.b is the rank 0 mon.


Related issues 3 (0 open3 closed)

Has duplicate RADOS - Bug #38893: RuntimeError: expected MON_CLOCK_SKEW but got noneResolved03/22/2019

Actions
Copied to RADOS - Backport #40228: nautilus: mon: rados/multimon tests fail with clock skewResolvedNathan CutlerActions
Copied to RADOS - Backport #44908: mimic: mon: rados/multimon tests fail with clock skewResolvedNathan CutlerActions
Actions

Also available in: Atom PDF