Actions
Bug #19753
closedDeny reservation if expected backfill size would put us over backfill_full_ratio
Status:
Resolved
Priority:
High
Assignee:
David Zafman
Category:
Backfill/Recovery
Target version:
-
% Done:
0%
Source:
Tags:
Backport:
Regression:
No
Severity:
3 - minor
Reviewed:
Affected Versions:
ceph-qa-suite:
Component(RADOS):
OSD
Pull request ID:
Crash signature (v1):
Crash signature (v2):
Description
We currently just check the full status based on disk usage. We need to adjust for the amount of data expected less the data already present.
Need to add the pg stat total in the reservation request?
Updated by Greg Farnum almost 7 years ago
- Project changed from Ceph to RADOS
- Category set to Backfill/Recovery
- Component(RADOS) OSD added
Updated by David Zafman almost 6 years ago
- Status changed from New to In Progress
Updated by David Zafman almost 6 years ago
- Blocked by Bug #24801: PG num_bytes becomes huge added
Updated by David Zafman over 5 years ago
Updated by imirc tw about 5 years ago
during backfilling after a failed disk the log files get spammed with do_backfill messages. log files easily grow beyond several GiB within hours
this is caused by https://github.com/ceph/ceph/pull/22797/files#diff-fb41013d27e932534adb50eb3de2aaa5R4273
is this expected behavior or is the log level on the dout too low?
Updated by David Zafman about 5 years ago
A couple of dout(0) should be dout(20) or some dout(10) for some less repetitive ones.
Updated by imirc tw about 5 years ago
would save a lot of diskspace if you could fix that :)
Updated by David Zafman about 5 years ago
- Related to Bug #39383: Too much log output generated from PrimaryLogPG::do_backfill() added
Actions