On 09/13/2013 06:20 AM, Sean Dague wrote: > Tempest has had a set of tests called whitebox that came in late in the > Folsom cycle. The idea being that you gave tempest db creds, and these > tests would check for the db being in a certain state after certain API > calls. It would also inject certain database state to get specific API > results. This was helpful in nailing some state races at the time. > > But since then these have largely bitrotted, especially with the db > schemas moving as fast as they do. As we started trying to make sure > everything in the Tempest tree was actually running regularly these were > failing a lot. David Kranz and I recently looked at them, and what we > had seemed to be largely Nova state change tests (doesn't affect any > other service). > > Going parallel made this worse, because isolation against the db is an > entirely different kind of problem than tenant isolation. > > So, that being said the call we made on IRC and in the QA meeting is to > get rid of them, and instead port the checks that we had into Nova unit > tests (which currently covered about 50% of the state checks that we in > whitebox). This brings it closer to the project which seems the right > approach. > > We are tracking this with - https://bugs.launchpad.net/tempest/+bug/1222852 > > The Tempest half is now done, the Nova half has a volunteer. > > While we are pretty sure that we got everyone's take on this, now is the > time to raise final objections to the approach.
I think this is an excellent plan. _______________________________________________ OpenStack-dev mailing list [email protected] http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
