Ceph » rbd

This can lead to a race if the snapshot is the parent for a clone and the clone starts to replicate (thus attaching itself to the parent). This child will prevent the snapshot copy logic from unprotecting/deleting the (unknown) snapshot since it has a new child. Since the clone logic looks up the parent snapshot by name in rbd-mirror, perhaps the snapshot create logic can utilize a dummy snapshot name upon creation, update the snap_seq mapping on disk, and then rename the snapshot to the correct name.