This is the schedule for the Kilo Design Summit, where OpenStack contributors discuss the future of OpenStack development. Click here for the main OpenStack Summit conference schedule.
Our current methodology of using tempest to house functional tests which are run against all projects on every commit has several problems * It puts the testing focus for all projects on the tempest team which is not scalable * There is not time to run deep functional tests because too much time is spent running tests against projects that cannot be affected The move to per-project functional testing, combined with more aggressive post-merge testing, will make catching bugs per-commit more likely and reduce the chance that bad code slipping in to some project will impact the gate stability for other projects. Some issues that need resolution: * How do we retain the ability to export a functional test suite for OpenStack as a whole, the way tempest is used now? This is important for refstack and any one else doing validation. * What kinds of non-scenario, non-integration tests remain in tempest? ** None? ** Some coverage for API stability? * How do we make sure projects do their testing in a consistent way? Do they need to? * Where do two-project integration tests live?