Project

General

Profile

Bug #43716

radosgw not resharding bucket properly

Added by Chris Durham 5 months ago. Updated 5 months ago.

Status:
Won't Fix
Priority:
Normal
Assignee:
-
Target version:
% Done:

0%

Source:
Community (dev)
Tags:
radosgw reshard
Backport:
Regression:
No
Severity:
2 - major
Reviewed:
Affected Versions:
ceph-qa-suite:
Pull request ID:
Crash signature:

Description

Using release 9002 on el8

Multisite configuration with two zones. Replication working fine with a given bucket

Add 300K objects to bucket. radosgw-admin bucket limit check shows "num_shards": 0 and fill_status: OVER 100%000, "num_objects:" 300000

over a weekend, the bucket never gets auto resharded, which I would expect it to do

Running radosgw-admin bucket reshard --bucket=bucketname --num-shards=32 properly reshards bucket on the master side

Replicated side radosgw-admin bucket limit check shows "num_shards": 32, but "num_objects:" 0. a listing of the pool itself still show the underlying objects, but using the aws s3api shows no objects

radoswgw-admin metadata get bucket:bucketname shows the same bucket instance id on both master and replicated side

If I then do a radosgw-admin bi list --bucket=bucketname on master, save the file, and move it over to the replicated site, then do a radosgw-admin bi put --bucket=bucketname < file

all the objects show up now with the s3 api

Note; I couldn't figure out how to do the bi put with the whole index file. doing it one object at a time works though...

History

#1 Updated by Chris Durham 5 months ago

after doing radosgw-admin bi put for the objects on the replicated side, I do a radosgw-admin bucket check --fix --check-objects --bucket=bucketname and afterwards radosgw-admin bucket limit check shows the proper number of objects again

#2 Updated by Casey Bodley 5 months ago

  • Status changed from New to Won't Fix

resharding in multisite isn't supported, and neither is the workflow of copying bi entries between sites

Also available in: Atom PDF