Comment 4 for bug 1550639

Another issue that worries me is nova doesn't change the task_state on the instance during a swap_volume operation, so the owner of the server instance could delete it while volumes are being swapped, which could leave the volumes in a garbage state on the cinder side. This is especially bad because swap_volume is by default an admin-only API, and it's initiated from cinder when doing a volume migration. So the user might not even know their volume backends are being migrated, and they decide they no longer need the instance and go to delete it, and that fails, or it passes but the volumes are stuck or orphaned in cinder.