Volume IO usage gets reset to 0 after a reboot / crash

[Michael Kerrin]

There is a lose of information in the volume usage after a reboot / crash of an instance.

[Michael Kerrin (michael-kerrin-w)]

To reproduce:

Watch volume.usage notifications.

Perform IO on volume and see notifications increase overtime.

Reboot the instance.

volume.usage notifications have been reset to 0.

The cumulative statistics about the IO performed in step 2 is lost. Customers will be undercharged on IO requests.

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/26993
Committed: http://github.com/openstack/nova/commit/a4cb456631f0e70685fbc361eced9baa477a8c56
Submitter: Jenkins
Branch: master

commit a4cb456631f0e70685fbc361eced9baa477a8c56
Author: Michael Kerrin <email address hidden>
Date: Tue Apr 16 09:43:28 2013 +0000

    Volume IO usage gets reset to 0 after a reboot / crash

    Check for reset of block device stats before updating volume usage cache.

    If an instance is reboot or crashes, the results of block device stats for
    that domain get reset. Add check that if block device stats are lower than
    the current stats in the database, then add current values to totals and
    update.

    Change-Id: I3011872f8d80a1519771f8423001650dcbcc7f5b
    Fixes bug: 1169097

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: stable/grizzly
Review: https://review.openstack.org/30130

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/30130
Committed: http://github.com/openstack/nova/commit/03c2429c3cf35c09941888c74d1a8dd47f37ad4e
Submitter: Jenkins
Branch: stable/grizzly

commit 03c2429c3cf35c09941888c74d1a8dd47f37ad4e
Author: Michael Kerrin <email address hidden>
Date: Tue Apr 16 09:43:28 2013 +0000

    Volume IO usage gets reset to 0 after a reboot / crash

    Check for reset of block device stats before updating volume usage cache.

    If an instance is reboot or crashes, the results of block device stats for
    that domain get reset. Add check that if block device stats are lower than
    the current stats in the database, then add current values to totals and
    update.

    Change-Id: I3011872f8d80a1519771f8423001650dcbcc7f5b
    Fixes bug: 1169097
    (cherry picked from commit a4cb456631f0e70685fbc361eced9baa477a8c56)