> > - parallelise dtests (because 12 hours is wild) > > That's one word for it. :) > > We used to ad hoc take a crack at sorting the individual test times by > longest and taking top-N and seeing if there was LHF to shave off that. > Being on a flight atm, not having that data handy right now, and that not > being in the linked logs from that pipeline run here (awesome work btw!), > do we think that might be something worth doing periodically on the project?
Yes I think so! Maybe even the longest dtest(s) can be an entry in the progress report? Especially now we can rewrite dtests into either quick "unit" tests using jvm-dtests or event diagnostics. Along we focus on dtests execution time, it would be nice to shore up the flakey unit tests (there's only a handful), so that they are more steps in the pipeline that hard fail (and fail-fast), giving faster feedback to the contributor/reviewer. --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@cassandra.apache.org For additional commands, e-mail: dev-h...@cassandra.apache.org
