Comment 30 for bug 1731349

Reviewed: https://review.openstack.org/531931
Committed: https://git.openstack.org/cgit/openstack/heat/commit/?id=32ec5141de1a8017ccfddc5dccf114deff8271ba
Submitter: Zuul
Branch: master

commit 32ec5141de1a8017ccfddc5dccf114deff8271ba
Author: Zane Bitter <email address hidden>
Date: Mon Jan 8 17:23:12 2018 -0500

    Don't load nested stack to get ResourceGroup blacklist

    In ResourceGroup, we (unwisely in retrospect) allow blacklisting of
    resources by reference ID (i.e. physical resource ID in most cases). In
    order to both determine whether a removal policy entry was an existing
    resource name and (if not) to find the resource by physical ID, we would
    always load the nested stack into memory.

    To avoid this, unconditionally create an output in the nested stack that
    returns a mapping of all the resource reference IDs and use it for finding
    a resource with a matching reference ID. Only if the output is not
    available (because the nested stack hasn't been updated since prior to the
    output being defined), fall back to the old way of doing it.

    Avoid looking up IDs for names that already appear in the blacklist
    (because they were blacklisted in a previous update and have been removed
    from the stack.)

    Since this is more expensive in time (though hopefully cheaper in space),
    update the blacklist only once on resource creation and whenever it
    actually changes during an update. This is accomplished by moving the
    update to a separate function that is called explicitly, rather than making
    it a side-effect of getting the current blacklist (which occurs up to 4
    times in each create/update).

    Change-Id: Ic32d7d99bad06f92b3d2919d58cd1f9128bcfe83
    Partial-Bug: #1731349