Project

General

Profile

Bug #34307

radosgw: dynamic resharding does not drop old bucket indexes

Added by Abhishek Lekshmanan 7 months ago. Updated 6 months ago.

Status:
New
Priority:
Normal
Target version:
-
Start date:
08/28/2018
Due date:
% Done:

0%

Source:
Tags:
Backport:
Regression:
No
Severity:
3 - minor
Reviewed:
Affected Versions:
ceph-qa-suite:
Pull request ID:

Description

ceph mon now complains when omap numbers exceed a threshold (12.2.6+, master/mimic), which can easily happen with large bucket indexes. From luminous we have dynamic resharding which while shards the bucket indexes, still seems to keep the old bucket indexes around, which have the large index which pg deep scrub will report. Have a radosgw-admin option to list (and maybe clear) these buckets and indexes so that they may be easily removed by the administrator.

History

#1 Updated by Abhishek Lekshmanan 7 months ago

bash oneliner find these bucket indexes using rgw admin commands

$ radosgw-admin bucket stats > bucket-stat.out
$ grep -E '"bucket"|"id"|"marker"' bucket-stats.out | awk -F ":" '{print $2}' | while read -r bucket; do read -r id; read -r marker; [ "$id" == "$marker" ] && true || echo $bucket $id $marker

#2 Updated by Matt Benjamin 7 months ago

I thought that the design intent of resharding was that old indexes would be retired once resharding was successfully completed. It seems that behavior should at least be an option, and probably the default?

Matt

#3 Updated by Matt Benjamin 7 months ago

  • Assignee set to Orit Wasserman

#4 Updated by Vikhyat Umrao 6 months ago

Orit had created this bug - http://tracker.ceph.com/issues/24082 I hope it is related to dynamic sharding.

Also available in: Atom PDF