I looks like "limit check" counts _all_ objects, but the resharder only uses stats from rgw.main. Is that correct?
If so, why? Shouldn't the auto-resharding act on the total number of objects in the bucket? I have around 5 million objects in rgw.none and 3 mil in rgw.main... limit check reports 8-9 million objects. (ceph version 19.2.3) Den man. 8. dec. 2025 kl. 10.10 skrev Johan Thomsen via ceph-users < [email protected]>: > Hi > > For some reason, the auto-resharding is not doing anything, even though it > reports fill status over 100%. See below: > > $ radosgw-admin bucket limit check --rgw-zonegroup=k1-objectstore > --rgw-zone=k1-objectstore > > > 2025-12-08T08:45:57.143+0000 7f8921528900 0 ERROR: current period > 43e92ddb-d0d2-4fbc-b18b-dc87126a7664 does not contain zone id > 1da0f602-9397-4997-8e76-3e3425468746 > > 2025-12-08T08:45:57.166+0000 7f8921528900 0 period > (43e92ddb-d0d2-4fbc-b18b-dc87126a7664 does not have zone > 1da0f602-9397-4997-8e76-3e3425468746 configured > > [ > > { > > "user_id": "k1-ceph-user", > > "buckets": [ > > { > > "bucket": "k1-snapshots", > > "tenant": "", > > "num_objects": 5766564, > > "num_shards": 97, > > "objects_per_shard": 59449, > > "fill_status": "OVER 114%" > > } > > ] > > } > > > Here's a dump of related config: > > > rgw_dynamic_resharding true > > rgw_max_dynamic_shards 1999 > > rgw_max_objs_per_shard 50000 > > rgw_md_log_max_shards 64 > > rgw_objexp_hints_num_shards 127 > > rgw_override_bucket_index_max_shards 1031 > > rgw_reshard_batch_size 64 > > rgw_reshard_bucket_lock_duration 360 > > rgw_reshard_max_aio 128 > > rgw_reshard_num_logs 16 > > rgw_reshard_thread_interval 600 > > rgw_safe_max_objects_per_shard 52000 > > rgw_shard_warning_threshold 90.000000 > > rgw_usage_max_shards 32 > > > Does anyone have an idea why it is not auto-resharding to a value below at > least 52k objs per shard? > _______________________________________________ > ceph-users mailing list -- [email protected] > To unsubscribe send an email to [email protected] > _______________________________________________ ceph-users mailing list -- [email protected] To unsubscribe send an email to [email protected]
