Hi again, Any thoughts / advice on the below, please? No obvious sign of any progress despite an entire weekend having gone by... On 22/05/2025 12:09, Matthew Vernon wrote:

Hi,

I'm running reef / cephadm. A user wanted two buckets removed (one of which was empty, and one of which had a large number (~half a million) of small objects in). In the primary zone I ran:

radosgw-admin bucket rm --bucket=name --bypass-gc --purge-objects [which I've previously used for removing buckets with very-many-objects in single-site just fine]

That returned OK for both buckets (quickly for the empty one, after about 20m for the many-objects one). 24 hours later, the empty bucket is gone in both zones, but the many-object one is still present in the secondary zone:

  source bucket :gitlab- artifacts[64f0dd71-48bf-45aa-9741-69a51c083556.75705.1])                 incremental sync on 11 shards                 bucket is behind on 11 shards                 behind shards: [0,1,2,3,4,5,6,7,8,9,10]

...which I think is showing that no shards have caught up, even a day later? Similar, sync status shows being behind on 1 metadata shard, and 11 data shards. There are no reported errors in sync.

Is there a way to see if this is actually making headway? And/or to get rid of the bucket in the secondary zone?

For future reference, would I have been better off trying to stop sync of the bucket and then run a bucket rm in both zones separately?

FWIW, this is a small (3-4 OSD hosts per site) and slow (hdds, with nvme for block.db) cluster.

Thanks,

Matthew