I think it is worth it to port Nicolas' enhancement for zombie test detection as a script that we can run after each build.
In my experience, cluster shutdown might take longer which resulted in hanging jvm. Cheers On Dec 24, 2012, at 8:36 AM, lars hofhansl <[email protected]> wrote: > This is looking at bit better now (I think I fixed TestReplication and > TestSplitTransactionOnCluster, making them less likely to fail) > > There two consecutive runs where no tests failed in the normal 0.94 build: > https://builds.apache.org/job/HBase-0.94/662/ and > https://builds.apache.org/job/HBase-0.94/661/ > > Unfortunately something hangs afterwards, so that the overall runs still fail > despite no failing tests. > If somebody knows how determine the hanging tests that would be great. > > Ted, I think you have some way to do that, right? > In trunk HadoopQA identified TestFromClientSide and TestRestartCluster as > "zombie tests", but of course in 0.94 there might different issues. > > -- Lars > > From: lars hofhansl <[email protected]> > To: "[email protected]" <[email protected]> > Sent: Saturday, December 22, 2012 2:59 PM > Subject: Re: 0.94 tests in a sorry state > > I created a new branch: 0.94-test and a new jenkins build for this branch: > hbase-0.94-test-tmp to track down these failures. > > > ________________________________ > From: lars hofhansl <[email protected]> > To: hbase-dev <[email protected]> > Sent: Friday, December 21, 2012 10:17 PM > Subject: 0.94 tests in a sorry state > > If anybody has some time to spend to check out some of the failing tests > that'd be very helpful. > https://builds.apache.org/job/HBase-0.94/ > > (I'm looking at TestReplication, which has been failing pretty reliably for a > while). > > > -- Lars >
