openstack-volume team mailing list archive

["Leahy, Oliver" <oliver.leahy@xxxxxx>]

    Hi,

    I reported bug #888649 on the Nova launchpad last week. Briefly, when a
    snapshot is created on a volume, and the volume is deleted, then the
    snapshot can never be deleted.

    Obviously that behavior is broken, so we're looking at fixing it in our
    volumes implementation. But we want to make sure that the behavior for our
    volumes and LVM based volumes is as far as possible the same, or if there
    are differences that they are necessary and logical.

    We can see three alternatives when nova-volumes receives a request to
    delete a volume that has been snapshotted :

    1. Block deletion of a volume if it is being referenced by a snapshot.
    2. Delete any snapshots referencing the volume and then delete the volume.
    3. Delete the volume and allow the snap shot to be deleted in the future.


    I think our preference is for option number 3 because our snapshots can
    exist happily after the source volume has been deleted. But I understand
    that this is not the case for LVM based volumes.

    Given this, what I propose doing is to have our volumes delete the volume
    and leave the snapshot available, and update the LVM code to block the
    deletion of a volume if it is being referenced by any snapshots.

    Is this the best way to proceed with this issue?