Restore to new volume ends up on the wrong host

[Duncan Thomas]

cinder/backup/api.py in the case of a restore to a new volume, creates the volume and then casts the restore request to the backup service. Unfortunately it casts it to the backup host name, not the volume host name, and since the lvm driver uses localpath in the restore code, this means the volume isn't found and the restore fails (lvm, multi-node case only)

i.e. the existing:
        self.backup_rpcapi.restore_backup(context,
                                          backup['host'],
                                          backup['id'],
                                          volume_id)

needs to be:
        self.backup_rpcapi.restore_backup(context,
                                          extract_host_from_volume(volume['host']),
                                          backup['id'],
                                          volume_id)

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/118709

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/118709
Committed: https://git.openstack.org/cgit/openstack/cinder/commit/?id=90bbed483618505620346255d0f1f3eeb539fea0
Submitter: Jenkins
Branch: master

commit 90bbed483618505620346255d0f1f3eeb539fea0
Author: Duncan Thomas <email address hidden>
Date: Wed Sep 3 19:07:14 2014 +0100

    During a restore send the restore request to the right host

    For restore from LVM volumes to work, the restore must be done
    on the host that serves the volume, since local_path is used
    in the restore code.

    Change-Id: Iad1fb5588b8365a0a165754389340bca3f1830b5
    Closes-Bug: #1365065