Fix #6676
osd should never block getting filestore throttle while holding the pg lock
Status:
New
Priority:
Normal
Assignee:
-
Category:
-
Target version:
-
% Done:
0%
Source:
other
Tags:
Backport:
Reviewed:
Affected Versions:
ceph-qa-suite:
Component(RADOS):
FileStore, OSD
Pull request ID:
Crash signature (v1):
Crash signature (v2):
Description
Currently, it is possible for the filestore finisher threads to block trying to acquire a pg lock in order to complete an op. This might cause that op to block until a filestore sync happens.
commit callback (filestore finisher)-> object recovery(filestore throttle)-> filestore sync
The filestore throttles should be modified so that the various OSD queues can throttle before dispatching an op based on a max op write size.
History
#1 Updated by Samuel Just over 10 years ago
- Target version deleted (
v0.73)
#2 Updated by Ian Colle over 10 years ago
- translation missing: en.field_story_points set to 5.0
#3 Updated by Patrick Donnelly about 5 years ago
- Project changed from Ceph to RADOS
- Start date deleted (
10/29/2013) - Component(RADOS) FileStore, OSD added