I wonder if this will be resolved by: https://github.com/ceph/ceph/pull/39358 <https://github.com/ceph/ceph/pull/39358> Deleting a bucket invokes an unordered listing, so that the objects in the bucket can be removed. There was a bug that caused this to loop back the objects. Eric
On Feb 23, 2021, at 3:56 PM, David Monschein <monschein@gmail.com> wrote:
Hi All,
We've been dealing with what seems to be a pretty annoying bug for a while now. We are unable to delete a customer's bucket that seems to have an extremely large number of aborted multipart uploads. I've had $(radosgw-admin bucket rm --bucket=pusulax --purge-objects) running in a screen session for almost 3 weeks now and it's still not finished; it's most likely stuck in a loop or something. The screen session with debug-rgw=10 spams billions of these messages:
...
We're running 14.2.16 on our RGWs and OSD nodes. Anyone have any ideas? Is it possible to target this bucket via rados directly to try and delete it? I'm weary of doing stuff like that though. Thanks in advance.
- Dave Monschein