live migrate of stopped VM goes to error, not back to task_state=None

[John Garbutt]

When the InstanceNotRunning exception is raised during the initial live-migrate, it should really just revert to task_state=None in the same way as if there was NoValidHost, or the compute service was down.

[Openstack Gerrit (openstack-gerrit)]

Fix proposed to branch: master
Review: https://review.openstack.org/90063

[John Garbutt (johngarbutt)]

Note this is related to, but different to:
https://bugs.launchpad.net/nova/+bug/1276214

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/90063
Committed: https://git.openstack.org/cgit/openstack/nova/commit/?id=d0a5e27288940c8e104330d784fed43575187afd
Submitter: Jenkins
Branch: master

commit d0a5e27288940c8e104330d784fed43575187afd
Author: John Garbutt <email address hidden>
Date: Thu Apr 24 12:44:26 2014 +0100

    Ensure live-migrate reverts if server not running

    When the check to ensure a VM is running before attempting the
    live migrate fails, it raises InstanceNotRunning. It should really be
    treated in the same way as MigrationPreCheckError.

    Along with fixing this bug, I have ensured the expected exceptions in
    migrate_server are in sync with the except clause in _live_migrate.

    This goes through to the API, ensuring the errors are not treated as
    internal server errors, but are given a slightly more appropriate error
    return value.

    Currently the API returns 400 for all these kinds of errors. While other
    error codes may be more appropriate, this change is concentrating purely
    on ensuring we treat all the exceptions in the same way.

    Change-Id: Iafe5026ef5a095d76e9864b489571e7bcfeb55f0
    Closes-Bug: #1312132