High numbers of revocation events cause excessive cache churn

[Morgan Fainberg]

Due to the explicit refresh on the revocation list cache, when a revocation event occurs there is an added extra load to the Keystone processes. This should be changed to an invalidate of cache to ensure that we do not have to regenerate the revocation_list except on actual demand. This reduces overall lock contention when using memcache / redis / external storage as the cache layer.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/45806

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/45806
Committed: http://github.com/openstack/keystone/commit/a9a1e1a4254c76f5f659af370dd6a1f1db1aac1a
Submitter: Jenkins
Branch: master

commit a9a1e1a4254c76f5f659af370dd6a1f1db1aac1a
Author: Morgan Fainberg <email address hidden>
Date: Mon Sep 9 19:10:01 2013 -0700

    Reduce churn of cache on revocation_list

    The explicit call to refresh will cause higher-than-average load
    under high counts of revocation events. The simple invalidate ensures
    that we don't re-generate the revocation list on revocation event
    but only on-demand.

    closes-bug: 1220359
    Change-Id: I826778843e344be001bbcc832054a0ad446285e7