Repeatedly resolving attributes causes high load

[Steve Baker]

Some resource attributes make API calls, so a template which calls the same attribute several times results in extra unnecessary requests. For a tripleo scaled overcloud this results in hundreds of unnecessary API calls plus long delays in resource creation and long periods of heat-engine 100% CPU usage.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/94755

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/94755
Committed: https://git.openstack.org/cgit/openstack/heat/commit/?id=ab412fee30d83816e38c57c837cf5c046f0790b5
Submitter: Jenkins
Branch: master

commit ab412fee30d83816e38c57c837cf5c046f0790b5
Author: Steve Baker <email address hidden>
Date: Thu May 22 13:24:38 2014 +1200

    Implement caching of resource attributes

    The default caching mode cache_local will return the previously
    resolved value if it is available.

    Whenever any resource changes state, all resource attribute caches
    are cleared just in case the state change has side-effects in
    other resources.

    The caching mode cache_none performs no caching, and is chosen
    for attributes which are one of the following:
    * Derived from any resource's metadata, resource data or resource_id
    * An API call which returns a secret

    Caching currently only exists for the duration of the parser.Stack
    object, but there is future potential for a caching mode which
    caches attributes spanning multiple requests.

    Closes-Bug: #1321970

    Change-Id: I01bf2983b726f0e81a2b8d5be94627353bdeb406