vm_state DELETED race in resource tracker

[Brian Elliott]

When an instance is deleted, the instance's vm_state is set to DELETED and then the actual DB record is deleted in separate DB calls.

If the resource tracker's audit process happens to run between the two database calls, resource tracker will erroneously report too much free usage.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/18835

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/18835
Committed: http://github.com/openstack/nova/commit/334d7b8d25b0aa9b41b578df8309497e34b0a05f
Submitter: Jenkins
Branch: master

commit 334d7b8d25b0aa9b41b578df8309497e34b0a05f
Author: Brian Elliott <email address hidden>
Date: Wed Jan 2 19:24:00 2013 +0000

    Fix race in resource tracker

    Fix a race in resource tracker where usage can get under reported if the
    audit process runs after an instance gets set to vm_state=DELETED, but
    before the DB record itself is deleted.

    bug 1095403

    Change-Id: I47943d12155a6484c6cb3adb31dfa4db0191cf26