Description of problem:

There is an issue where bilogs are left in the sync-disabled buckets in the multi-site sync-enabled cluster. In our cluster, these logs were accumulated for a long time and caused a performance problem.

This phenomenon can be reproduced through the following procedure:
First off, create multi-site cluster(c1,c2) using rgw/test-rgw-multisite.sh script. And then execute the below commands.

Version-Release number of selected component (if applicable):


How reproducible:


Steps to Reproduce:

$ yum install s3cmd -y
$ cat >> c1-s3cmd << EOF
[default]
access_key = 1234567890
secret_key = pencil
host_base = 127.0.0.1:8001
host_bucket = 127.0.0.1:8001
use_https = False
EOF
$ s3cmd -c c1-s3cmd mb s3://mytest1
$ ../src/mrun c1 radosgw-admin bucket sync disable --bucket mytest1
$ s3cmd  -c c1-s3cmd put obj s3://mytest1
$ ../src/mrun c1 radosgw-admin bilog list --bucket mytest1 | grep -A1 write

    (Nothing left)

$ s3cmd -c c1-s3cmd mb s3://mytest2
$ ../src/mrun c1 radosgw-admin bucket sync disable --bucket mytest2
$ ../src/mrun c1 radosgw-admin bucket check --fix --bucket mytest2
$ s3cmd  -c c1-s3cmd put obj s3://mytest2
$ ../src/mrun c1 radosgw-admin bilog list --bucket mytest2 | grep -A1 write
        "op": "write",
        "object": "obj",

    (a bilog is left in the sync-disabled bucket)

Actual results:


Expected results:


Additional info: