Project

General

Profile

Actions
Copy link

Bug #42590

open

Thrasher can set full ratio but no yaml whitelists for (OSD_OUT_OF_ORDER_FULL)

Added by David Zafman over 4 years ago. Updated over 4 years ago.

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

0%

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

Description

We have 2 choices:

Add (OSD_OUT_OF_ORDER_FULL) to the appropriate yaml files.

Or simpler yet set all values but failsafe full, so that full ratio isn't lower than nearfull and backfillfull.

self.ceph_manager.raw_cluster_cmd('osd', 'set-nearfull-ratio', '.001')
self.ceph_manager.raw_cluster_cmd('osd', 'set-backfillfull-ratio', '.002')
self.ceph_manager.raw_cluster_cmd('osd', 'set-full-ratio', '.003')

Should the thrasher get the previous values and then set them back again?

This is very rare because the probability of this trash is very low (.05)

Seen in luminous smoke test:

smoke/basic/{clusters/{fixed-3-cephfs.yaml openstack.yaml} objectstore/bluestore-bitmap.yaml
tasks/rados_bench.yaml}

failure_reason: '"2019-10-31 20:42:33.506775 mon.a mon.0 172.21.15.180:6789/0 1121
: cluster [ERR] Health check failed: full ratio(s) out of order (OSD_OUT_OF_ORDER_FULL)"
in cluster log'

Actions
Copy link
 #1

Updated by David Zafman over 4 years ago

  • Backport set to mimic, nautilus
Actions
Copy link

Also available in: Atom PDF