Unhandled exception on update_volume_group when vg is None

[John Griffith]

There's a possibility of the cinder-scheduler issuing a get_volume_stats() call before the volume-service/driver is properly initialized and ready. The result is an rpc Error and corresponding stack-trace in the cinder-volume logs.

We should check to make sure the VG is initialized and ready before attempting to update the stats, if it's not just return the empty stats dict and let the schedulers periodic update things when it's finally ready.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/45484

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/45484
Committed: http://github.com/openstack/cinder/commit/d552220a83052daabe19682b531bc6b8e685a834
Submitter: Jenkins
Branch: master

commit d552220a83052daabe19682b531bc6b8e685a834
Author: John Griffith <email address hidden>
Date: Fri Sep 6 13:55:39 2013 -0600

    Do not attempt vg.update on uninitialized vg

    It's possible for the scheduler to send the periodic
    stats update request *before* the volume service/driver
    has finished initializing, resulting in an exception trace
    being logged.

    This change adds a check to verify that the vg is actually
    initialized (completed check_for_setup_error) and it also
    removes all of the duplication of each LVM class driver.

    To deal with other drivers that may have similar issues,
    add initialized member to base driver, that member is
    set False on __init__ and is updated by the manager
    after running check_for_setup_error. We also skip
    update_stats if initialization isn't set True.

    Rather than have a separate copy of update_stats in every
    driver, we have one in the base LVM class and we set the two
    unique variables as member parameters.

    Fixes bug 1221874

    Change-Id: I159e98a77782b8b2c85a8dd956b150828358fd25