Metadata writes are inherently races

[Zane Bitter]

Calling Resource.metadata_set from multiple threads/processes is always a potential race condition. The potential for this is quite high compared to other writes to the Resource table, since such writes are routinely done outside of normal stack create/update operations, e.g. in resource_signal().

One instance of this race was fixed for bug 1477329, and others were papered over for bug 1488366 and bug 1542455 by not updating the metadata when we don't have to. Nevertheless, the race remains.

[OpenStack Infra (hudson-openstack)]

Related fix proposed to branch: master
Review: https://review.openstack.org/276917

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/276917
Committed: https://git.openstack.org/cgit/openstack/heat/commit/?id=40a41ec076df06bcdd290e7fae6140c4b2c40145
Submitter: Jenkins
Branch: master

commit 40a41ec076df06bcdd290e7fae6140c4b2c40145
Author: Zane Bitter <email address hidden>
Date: Fri Feb 5 15:08:01 2016 -0500

    Log all writes to resource metadata

    I believe we're updating this more often than we want (see bug 1542455) and
    that there are race conditions in play (bug 1477329 was one, but there are
    others - see bug 1542480). Having a debug log will help us be more aware of
    what's happening.

    Change-Id: I6deb09c2cdeedc42ae359708aa29eb699596953e
    Related-Bug: #1542480