Hi! Lately our CI pipelines have bit-rotted (again) to the point of being unhelpful. In PHOENIX-6537 I have tried to update and fix them. I think that I have fixed the worst of the flaky test cases, and all systematic problems, and the remaining failures are either environmental (connection refused, Jenkins worker disconnect, etc), or unfixed flaky test cases.
Things to be aware of: The new Failsafe plugin version (3.0.0-M5) aggregates @Category annotations on a class and its superclasses, and will try to run them in each corresponding execution (i.e. you cannot override @Category in a child class anymore). I have refactored the existing tests so that only the leaf classes have annotations. Make sure that you explicitly add the right @Category to all new test classes. The new Yetus 0.13 version uses the PR check status field (at the bottom of the GH PR page) to report its status, it no longer generates GH PR comments. I could not get Jacoco to run reliably, so I had to disable it in the CI. (See PHOENIX-5893,) I have also disabled building the shaded artifacts and assembly during tests, this should save close to an hour on a PR run. I'm going to keep an eye on the test results, and open tickets / push addendums for problems that crop up for the next while. Please help by checking the CI results and reporting (and hopefully fixing) both flaky test cases, and problems with the CI pipeline itself. At the moment the changes are live on master and 5.1, but I'm going to backport them to the 4.x branches soon. Regards Istvan
