> > Mick, did I miss anything? >
Yup, ci-cassandra.a.o needs to be our canonical CI because of the reasons you state, and because it's the only one fully usable by a volunteer based PMC, i.e. community donated hardware, controlled by ASF, and no need for a proprietary licence. I think we should be making that post-commit pipeline as comprehensive as possible. I also really appreciate that we have CircleCI, apart from its benefits (particularly speed for pre-commit), it provides valuable double-accounting over CI results. A few times we have broken because one system didn't properly catch errors. It can be frustrating that we get green in one system and then it breaks in the other, but if we embrace this strengthening tactic: only green in both counts as green; then we end up in a better place. Let's continue to make post-commit CI as comprehensive as possible, and figure out as-we-go what the most efficient pre-commit gate is, and accept that if you break it post-commit you still broke it. For releases I think we have to have no hard failures, and less flakies than the previous release. Using Butler we can more easily visualise failures from flakies from CI infra instabilities, and ensure tickets are created and if a dev broke it that they are assigned accordingly (and prioritise its fix above all other non-critical work in the project). --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@cassandra.apache.org For additional commands, e-mail: dev-h...@cassandra.apache.org
