+1 As far as the failure rate goes, even if the rate is 25% restarting the pipeline usually gives a passing run. And this would make us less likely to incorrectly dismiss a failure as not related. If the second build also fails with the same failures, those should be investigated and resolved in some way before the PR is merged.
On Fri, Nov 9, 2018 at 1:49 PM Anthony Baker <aba...@pivotal.io> wrote: > It looks like the current failure rate (post-PR, including all types of > failures) for DistributedTest is around 25%. Do most people experience > similar failure rates on the *-pr pipeline? I’m specifically wondering > about failures unrelated to your changes. > > Anthony > > > > On Nov 9, 2018, at 12:55 PM, Dan Smith <dsm...@pivotal.io> wrote: > > > > Hi all, > > > > Kirks emails reminded me - I think we are at the point now with our PR > > checks where we should not be merging anything to develop that doesn't > pass > > all of the PR checks. > > > > I propose we disable the merge button unless a PR is passing all of the > > checks. If we are in agreement I'll follow up with infra to see how to > make > > that happen. > > > > This would not completely prevent pushing directly to develop from the > > command line, but since most developers seem to be using the github UI, I > > hope that it will steer people towards getting the PRs passing instead of > > using the command line. > > > > Thoughts? > > -Dan > >
