Hi everyone, So as the title indicates I pushed through the change that migrates tempest to run by default in parallel. (except for the neutron job, which still needs work to run in parallel) This should improve the run times for tempest jobs to between 20-30min on average which will definitely help with the upcoming milestone rush. However, the short term trade-off of this is that it adds a bit more flakiness to the tempest runs. But, it's a good flaky albeit an annoying one. :)
Running in parallel actually helps us simulate a real workload a bit better than before. Since we're running 4 test cases at once it more closely resembles a real OpenStack deployment (at least compared to running the tests one at a time) Just in debugging things to get it ready for running in the gate, I uncovered some race conditions that pre-existed in the projects, and I'm sure there are some more. There are probably still races between tests in tempest as well that will have to be sorted. So I'm asking everyone to be diligent with watching (and using) the recheck list and that all the devs just don't recheck and forget. We need a combined effort from everyone to make things stable and debug seemingly random failures when they come up. If we can iron out most of the flakiness by the milestone rush having the faster runtime will definitely help make the surge easier for everyone. Thanks, Matt Treinish _______________________________________________ OpenStack-dev mailing list OpenStack-dev@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
