Multiple delete_on_terminate volumes may not get deleted

[Andrew Laski]

There's no catching of exceptions in _cleanup_volumes in compute/manager.py so a raised exception will short circuit cleaning up later volumes.

    def _cleanup_volumes(self, context, instance_uuid, bdms):
        for bdm in bdms:
            LOG.debug("terminating bdm %s", bdm,
                      instance_uuid=instance_uuid)
            if bdm.volume_id and bdm.delete_on_termination:
                self.volume_api.delete(context, bdm.volume_id)

[Andrew Laski (alaski)]

I seem to have bad luck with my patches updating LP bugs, but I have a proposal up at https://review.openstack.org/#/c/106879/ . I'm open to someone else taking over if they wish, but I have another fix which depends on this.

[Christopher Yeoh (cyeoh-0)]

Oh sorry Andrew - only took it because it said no one was assigned to it - its all yours :-)

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/106879
Committed: https://git.openstack.org/cgit/openstack/nova/commit/?id=70feb102ae86d3421e46a07710e4cbab35408d02
Submitter: Jenkins
Branch: master

commit 70feb102ae86d3421e46a07710e4cbab35408d02
Author: Andrew Laski <email address hidden>
Date: Mon Jul 14 17:01:34 2014 -0400

    Defer raising an exception when deleting volumes

    When looping through block device mappings to delete volumes with
    delete_on_terminate set the loop should not exit on an exception.
    Instead the loop will continue and the last exception can optionally be
    raised.

    Change-Id: If1c6ec30c6f49009ef42f9a3b3df1594783b0ae5
    Closes-Bug: #1341738