Intermittent "Delete snapshot failed, due to snapshot busy" on ceph in ceph gate

[John Griffith]

Saw this recently, looks like there's a few in logstash:

http://logs.openstack.org/93/212193/4/check/gate-tempest-dsvm-full-ceph/5d69556/logs/screen-c-vol.txt.gz#_2015-08-13_17_45_37_013

[Jordan Pittier (jordan-pittier)]

Yeah, I still see this a lot.

[Brin Zhang (zhangbailin)]

I think this is not a bug, Failed to delete the snapshot of the volume with rbd backend, because of the config of cinder-volume/cinder.conf there is a field "rbd_flatten_volume_from_snapshot".
The default value is False, when you create snapshots of a volume, will make a chain clone,cannot be deleted because you create a new volume or virtual machine with this snapshot.
On the other hand,if "rbd_flatten_volume_from_snapshot = True". When you create a volume of time will make a full copy,this is actually done with a flatten operation.So there is no dependency between this volume and the snapshot,and you can delete the snapshot.
If you want a deeper understanding of this operation, you can look at the create_volume_from_source interface until rbd driver.

I hope this helps you.