Project

General

Profile

Actions
Copy link

Bug #20690

open

Cluster status is HEALTH_OK even though PGs are in unknown state

Added by Nathan Cutler almost 7 years ago. Updated over 6 years ago.

Status:
Need More Info
Priority:
Normal
Assignee:
-
Category:
-
Target version:
-
% Done:

0%

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

Description

In an automated test, we see PGs in unknown state, yet "ceph -s" reports HEALTH_OK. The test sees HEALTH_OK and proceeds to mount cephfs, but this fails:

https://paste2.org/hj4yzUb5

This appears to be a regression introduced by https://github.com/ceph/ceph/pull/15643/

Related issues 1 (0 open — 1 closed)

Related to RADOS - Bug #20693: monthrash has spurious PG_AVAILABILITY etc warningsResolved07/19/2017

Actions
Actions
Copy link
 #1

Updated by Nathan Cutler almost 7 years ago

  • Project changed from Ceph to RADOS
Actions
Copy link
 #2

Updated by Nathan Cutler almost 7 years ago

  • Related to Bug #20693: monthrash has spurious PG_AVAILABILITY etc warnings added
Actions
Copy link
 #3

Updated by Nathan Cutler almost 7 years ago

This log excerpt illustrates the problem: https://paste2.org/cne4IzG1

The logs starts immediately after cephfs deployment ("Stage 4") finishes. Notice how 100+ PGs are in "unknown" state. Just 4 seconds later, all PGs are in "active+clean".

When the PGs are in "unknown" state, cephfs cannot be mounted.

Actions
Copy link
 #4

Updated by Sage Weil over 6 years ago

  • Status changed from New to Need More Info

why can't cephfs be mounted when pgs are unknown?

Actions
Copy link

Also available in: Atom PDF