the distributed test framework I was thinking of was grind/dist_test[1], thought we might need to tweak it if the surefire/failsafe distinction matters. we'd have to talk to asf infra about getting some VMs allocated for this, unless we can work out a way to launch them ourselves via jenkins.
also we can probably speed things up by grouping the ITs up and running sets of them against a shared clusterdock topology[2]. [1]: https://github.com/cloudera/dist_test/blob/master/docs/grind.md [2]: https://github.com/cloudera/clusterdock On Wed, Aug 10, 2016 at 10:25 PM, Sean Busbey <[email protected]> wrote: > Right now it is manually maintained. Specifically, updating it is a matter > of find + grep + copy/paste. > > There are some options for automating it, but I haven't seriously > investigated any of them yet. One would be using the job DSL stuff Jenkins > has to generate the job config. That should allow us to do the enumerating > as a part of a launching job. Another would be to use a distributed test > framework rather than Jenkins to do the parallelization. > > If no one else digs into these questions, I imagine I'll need to by the end > of the year for other work related stuff. > > -- > Sean Busbey > > On Aug 10, 2016 21:56, "Josh Elser" <[email protected]> wrote: >> >> For those who haven't noticed Sean's great work so far: >> >> https://builds.apache.org/view/A-D/view/Accumulo/job/Accumulo-master-IT/ >> >> Kudus to you and thanks for your work already! This is much better than I >> expected was possible on ASF infra. >> >> Sean, I do have one question for you: I see that the integration tests are >> provided as an "axis" to the test matrix (tests enumerated in job >> configuration). Is this something that we would need to update as more tests >> as added? Do you have plans (or something already in place) that could >> automatically figure out the set of ITs for the Jenkins job? >> >> - Josh -- busbey
