Actions
Bug #20847
closedlow performance for bluestore rbd block creation vs filestore
Status:
Closed
Priority:
Normal
Assignee:
-
Target version:
-
% Done:
0%
Source:
Tags:
Backport:
Regression:
No
Severity:
3 - minor
Reviewed:
Affected Versions:
ceph-qa-suite:
Pull request ID:
Crash signature (v1):
Crash signature (v2):
Description
It appears that bluestore may be slower than filestore when creating RBD blocks (ie if pre-filling an RBD volume with 4MB writes). This is most apparent when using a HDD block partition with the WAL and DB on flash. Bluestore however, appears to be faster in a similar setup when creating new 4MB objects via rados bench and when doing 4MB writes to existing RBD blocks.
I'll see if I can run some additional tests to capture with high logging enabled.
Mark
Updated by Mark Nelson over 6 years ago
- Assignee set to Sage Weil
- Priority changed from Normal to Urgent
Updated by Sage Weil over 6 years ago
- Related to Bug #20842: Ceph 12.1 BlueStore low performance added
Updated by Josh Durgin almost 5 years ago
Mark - is this fixed by deferred writes now?
Actions