Optimize periodic call to get_by_host

[Ed Leafe]

A recent change (https://review.openstack.org/#/c/160513) added a periodic task to the compute manager that retrieves a list of instances. During review it was suggested that the call could be optimized, but that it could be done as a later bugfix.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/165849

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/165849
Committed: https://git.openstack.org/cgit/openstack/nova/commit/?id=212f81e12443a98101e5987a1c308b00f1cacd76
Submitter: Jenkins
Branch: master

commit 212f81e12443a98101e5987a1c308b00f1cacd76
Author: EdLeafe <email address hidden>
Date: Thu Mar 19 14:23:19 2015 +0000

    Optimize periodic call to get_by_host

    This is an optimization suggested by John Garbutt in the review of
    https://review.openstack.org/#/c/160513

    To quote John:

    """
    We are close to the wire, so I don't want to block on the little nit I found, I
    think a follow up patch could fix that. But lets see what other folks think.

    The rest of it seems OK, I am just glad to see it clearly optional, I suppose.

    Most periodic tasks do this:

    db_instances = objects.InstanceList.get_by_host(
    context, self.host, expected_attrs=[], use_slave=True)

    But maybe lets just raise a bug and fix that in a follow on patch. Not sure
    what stuff you need merging in for the sync, you might need the flavor in
    system_metadata I guess, but we could add use_slave I think.
    """

    Change-Id: I998b6ac0ad6fd53769061ef66b9b19ba5e12391d
    Closes-Bug: #1434112