LinuxCon + CloudOpen + ContainerCon NA 2015

OpenStack is incredibly complex with tons of moving parts. As part of the community's continuous integration system, often called the gate, we run tests against a running OpenStack cloud. However, sometimes things fail for reasons unrelated to the patch being tested. These failures, often called "flakey failures" by developers, are normally an exposure of interesting timing problems in OpenStack and they can often be very tricky to debug. However, we've found that almost all of these issues and the difficulty in debugging them also apply to real deployments. This talk will go over the mechanics of debugging gate failures. As part of the discussion comparisons will be drawn to show how similar the process of debugging gate failures is to production OpenStack issues. To reinforce that anything we see in the gate most likely is being dealt with by users and operators.