Nova compute base image cleanup

We would like to have the possibility to delete the base image from all nova-compute nodes when deleting an image from the glance server without having to rely on an expiration timer. (remove_unused_original_minimum_age_seconds)

For instance "glance purge image-id" could be used in order to delete the image from glance and then send an AMQP message in order to delete that ID from all nova-computes.

If that isn't possible, perhaps delete it from the computes where an instance was deployed using that image.

That command should be allowed to be run by an admin account only in order to avoid caching/decaching heavily used images.

Perhaps the image deletion depending on a timer should be replaced by a DB call to see if it's deleted. If the image is deleted from glance and nobody uses the base image any longer, there's
no reasons to keep it on disk ...

If the later idea is implemented, perhaps a kind of trigger could be implemented (90% of /var/lib/nova/instances/_base is used) before considering deleting cached images.

Keeping deleted images on a compute node can lead to DoS.