Thanks for taking this on Mick. We had roughly 7 jenkins slaves DS had donated languish I'm looking into (came up on slack yesterday morning) so hopefully we'll have some more resources back in the pool soon.
One other thing we've seen repeatedly over the years is startlingly low-hanging-fruit of some badly performing tests (long stress generation, hangs leading to timeout kills, etc) that greatly extend the testing duration beyond what's strictly necessary to get the signal out of it. There may be value in further conversation about a periodic "health check" on tests and pruning out / refactoring / cleaning up some of the most egregious offenders in terms of inefficiency over time. Also happens to be a really good way to ramp on the code-base; for me, working through rewriting unit tests for the StorageEngine rewrite definitely gave me a broad overview of things I don't think I'd otherwise have gotten. On Wed, Jan 8, 2020 at 12:34 AM Mick Semb Wever <m...@apache.org> wrote: > > I'm chuffed to say that ASF Jenkins builds for branches 2.2, 3.0, 3.11, > and trunk, are now all working. > > Trunk: https://builds.apache.org/view/A-D/view/Cassandra%20trunk/ > 3.11: https://builds.apache.org/view/A-D/view/Cassandra%203.11/ > 3.0: https://builds.apache.org/view/A-D/view/Cassandra%203.0/ > 2.2: https://builds.apache.org/view/A-D/view/Cassandra%202.2/ > > > The current setup for Cassandra's CI is a bit of a mess. Recently there's > been more attention on CircleCI, because it has shown to be less flakey > than ASF's Jenkins, and allows automated unit testing any contributor's > patches. Still CircleCI has short-comings: it provides no canonical build > status for our supported branches, you have to hack commit the circleci > yaml, and unless you work for a (super) large company that is willing to > give you 2000+ CircleCI containers the full test suite is out of your > reach, not cool for an OSS project. (If I'm wrong on that last point please > correct me!) > > So… I've gone against the grain and tried to give Jenkins a little love. > All the builds now are running again, except the dtest-large jobs which > require servers with 32GB+ ram which we no longer have. And the branches > each have their own view, as listed above. > > In addition there's also a view for custom builds to test patches > https://builds.apache.org/view/A-D/view/Cassandra%20%C2%A0patches/ > and in there is a pipeline job I'm working on so that patches only have to > be entered once: > https://builds.apache.org/view/A-D/view/Cassandra%20%C2%A0patches/job/Cassandra-devbranch-pipeline/ > > If people have tickets with patches that they would like to test, ping me > and I'll happily enter them. And anyone with multiple patches I'd > definitely recommend setting up your own Jenkins, as Stefan Podkowinski has > done a pretty awesome job at automating the Jenkins job creation: > http://cassandra.apache.org/doc/latest/development/ci.html > > One of the things I'd like to do next, as jobs stabilise, is to continue > to add email failure notifications so we're quicker to catch regressions on > our main branches. And I'm curious as to what others feel is needed and > should be the direction we head in. > > regards, > Mick > > > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: dev-unsubscr...@cassandra.apache.org > For additional commands, e-mail: dev-h...@cassandra.apache.org > >
