Only "TestRackawareEnsemblePlacementPolicyUsingScript" failed for me. But I suspected it might be related to my environment. will look into it.
Failed tests: TestRackawareEnsemblePlacementPolicyUsingScript.testNewEnsembleWithEnoughRacks:320 expected:<3> but was:<2> - Sijie On Wed, Nov 8, 2017 at 9:21 AM, Sijie Guo <guosi...@gmail.com> wrote: > > > On Wed, Nov 8, 2017 at 7:45 AM, Ivan Kelly <iv...@apache.org> wrote: > >> > This has been done deliberately. Like for 4.5.0. >> > I am ok if you PMC want to do it differently. >> > Or maybe I have misunderstood the guide. >> > You already stated this on slack. I apologize I went forward following >> the >> > guide >> Oh, I'm misremembering how we used to do it, another project I worked >> on had it different. >> >> However, the point still stands. The final release should have a >> different sha1 than the candidates, otherwise people's maven caches >> get messed up, leading to hidden problems down the line. > > > Are you talking about a tag with "-rc" or artifacts with "-rc"? > > I don't think we've ever used "rc" for the staging artifacts. if you take > a look at the "staging repositories" at nexus, > I don't see any projects using "rc" for the staging artifacts. > > The purpose of a staging area is once a RC is approved, you can just click > release to push it to stable and it will be synced to maven central. > There should be no other changes in between. > > You can read the ASF guide about "publishing maven artifacts" - > http://www.apache.org/dev/publishing-maven-artifacts.html > > However, I do see different projects have different convention about the > rc tag name. some projects are using the final tag names and dropping > the tag if a rc doesn't pass; some projects are using tag names with > "-rc<rc_num>", and clean up the tag names after an RC is approved. > We have been always using the final tag names for releases, but if a rc > tag is preferable, we can go down in that route. > > > >> >> >> I ran the tests twice, and both times the following failed for me. >> >> They're probably flakes, but they failed on 2/2 so they should be >> >> investigated. This is on a machine that sometimes runs master tests >> >> cleanly (we really need to sort out our flakes). >> >> - BookKeeperAdminTest.testDecommissionBookie >> >> - >> >> TestRackawareEnsemblePlacementPolicyUsingScript.testNewEnsem >> bleWithEnoughRacks >> >> - BookKeeperDiskSpaceWeightedLedgerPlacementTest >> >> >> > >> > Ok let's dig into this. >> It would be good for others to confirm whether they see the same >> failures, even if only periodically. >> > > verifying the build on my laptop. will update the results here once it is > done. > > >> >> -Ivan >> > >
