+1 to allowing this PR to be merged, although I'd lean strongly toward facilitating this by temporarily increasing the timeout on the job to allow it to actually pass rather than a manual override of the StressNewTest.
The fact that it's passed over 7000 times without failing is pretty strong evidence that it's not a flaky test, which is what StressNewTest is supposed to catch, so there doesn't seem to be any risk associated with circumventing it in this case, but if there's a feasible solution that doesn't involve "cheating" or ignoring the test job, then that would be preferable. - Donal On Thu, Oct 31, 2019 at 2:04 PM Jason Huynh <jhu...@pivotal.io> wrote: > Greetings, > > We have a pull request (https://github.com/apache/geode/pull/4250) that is > running into a problem with stressNewTest. Mostly the tests that are being > run are RollingUpgrade tests that take quite a bit of time to run the full > suite. Because these tests are added/modified, the stressNewTest doesn't > have enough time to complete the run because it runs them N(50) number of > times. > > However what has completed is 7400 tests and none of them have failed: > > http://files.apachegeode-ci.info/builds/apache-develop-pr/geode-pr-4250/test-results/repeatTest/1572546653/ > > We would like to get this fix in before branching the next release, but are > unable to due to stressNewTest gating the merge button. I know we have > another thread about overrides etc, and maybe this is a data point, but > this isn't meant to discuss that. > > Would everyone be able to agree to allow someone to manually override and > merge this commit in (title of PR and reviews pending)? >
