Ceph » rgw

radosgw send data-log list infinitely when opposite end trimmed the data-log and in quiescence.

the sending side always send data-log list with this log:
2017-08-08 16:59:01.979654 7f7a11fe3700 20 data sync: incremental_sync:1340: shard_id=63 datalog_marker=1_1502160434.853416_118147.1 sync_marker.marker=

And in the opposite end, the related data log has already been trimmed.

when doing rados listomapkeys p master.rgw.log data_log.63, no output.
And the omapheader of data_log.63 is: (1_1502160434.853416_118147.12&Y

Sending data-log infinitely caused the opposite end stuck with high cpu usage.

Can you provide the rgw logs?

Orit Wasserman wrote:

Can you provide the rgw logs?

useful info of the rgw logs is:
2017-08-08 16:59:01.979654 7f7a11fe3700 20 data sync: incremental_sync:1340: shard_id=63 datalog_marker=1_1502160434.853416_118147.1 sync_marker.marker=

I have opened a PR to fix this issue:
https://github.com/ceph/ceph/pull/16926