Hello A few weeks earlier I've tried to hunt down this problem... so...to my best knowledge the cause of this seems to be the following:
* in some cases the "cleanup" after a failed query may somehow leave some threads behind... * these threads have reference to the "customized" session classloader - this makes the threads more memory hungry * after a while these threads/classloaders eat up the heap... I've opened HIVE-18522 for this thread issue I think this problem is not new ...and it might have been present earlier as well...the only thing what changed is that there were a few more new features which have added new udfs/etc which made the memory cost of a session more heavier.. ...and as a sidenote: I'm not convinced that this issue will arise in a proper hs2 setup - as it might be easily connected to the fact that these tests are using the cli driver to execute the tests. cheers, Zoltan On 7 Apr 2018 7:15 p.m., Ashutosh Chauhan <hashut...@apache.org> wrote: We need to investigate and find out root cause of these failures. If its determined that its a corner case and fix is non-trivial then we may release note it under known issues. But ideally we should fix these failures. Cutting a branch should make it easier since branch is expected to receive lot less commits as compared to master so it should be faster to stabilize branch. On Fri, Apr 6, 2018 at 10:49 AM, Eugene Koifman <ekoif...@hortonworks.com> wrote: > Cutting the branch before the tests are stabilized would mean we have to > fix them in 2 places. > > On 4/6/18, 10:05 AM, "Thejas Nair" <thejas.n...@gmail.com> wrote: > > That needs to be cleaned up. There are far too many right now, its > just not handful of flaky tests. > > > On Fri, Apr 6, 2018 at 2:48 AM, Peter Vary <pv...@cloudera.com> wrote: > > Hi Team, > > > > I am new to the Hive release process and it is not clear to me how > the failing tests are handled. Do we plan to fix the failing tests before > release? Or it is accepted to cut a new major release with known test > issues. > > > > Thanks, > > Peter > > > >> On Apr 5, 2018, at 8:25 PM, Vineet Garg <vg...@hortonworks.com> > wrote: > >> > >> Hello, > >> > >> I plan to cut off branch for Hive 3.0.0 on Monday (9 April) since > bunch of folks have big patches pending. > >> > >> Regards, > >> Vineet G > >> > >>> On Apr 2, 2018, at 3:14 PM, Vineet Garg <vg...@hortonworks.com> > wrote: > >>> > >>> Hello, > >>> > >>> We have enough votes to prepare a release candidate for Hive > 3.0.0. I am going to cutoff a branch in a day or two. I’ll send an email as > soon as I have the branch ready. > >>> Meanwhile there are approximately 69 JIRAs which are currently > opened with fix version 3.0.0. I’ll appreciate if their respective owners > would update the JIRA if it is a blocker. Otherwise I’ll update them to > defer the fix version to next release. > >>> > >>> Regards, > >>> Vineet G > >>> > >> > > > > > >
