On Mon, Aug 31, 2015 at 8:01 AM, Sean Busbey <bus...@cloudera.com> wrote:
> FYI, I've changed the precommit test to run using the test-patch.sh file
> from a branch for HBASE-14337 so I can make changes needed to allow the
> workspace to have a maven repo just for our builds (rather than the shared
> one with the jenkins user).
>
> submitted patches are still applied and run against their branch (default
> master).
>
> Once I have a coherent patch I'll post it to HBASE-14337 and we can stop
> using the branch.
>
>
I just ran into this. It took a bit to figure it ("Why are my changes to
test-patch.sh not having an effect..." -- smile).
So, I think its probably a good idea to run test-patch.sh from a branch.
The alternative is checking in latest test-patch.sh changes to all branches
(test-patch.sh and pieces of this script are scattered enough already with
the zombie finding portion pasted as after-build script to run on build
completion on a bunch of our branches -- need to fix).
I added a branch to our repo called jenkins.build. It should be clearer
that this is where we do jenkins.build from (a bit clearer than HBASE-14337
-- smile). I then changed the patch build to use this branch doing jenkins
build (added comments in the script up in jenkins so hopefully it is a bit
plain what is going on).
Good on you Sean,
St.Ack
> On Fri, Aug 21, 2015 at 2:17 PM, Stack <st...@duboce.net> wrote:
>
> > A deadlock was causing TestDistributedLogSplitting to hang. Elliott fixed
> > it over in HBASE-14274. This was the source of the zombie processes
> > reported by INFRA (at least, it matched the one instance they reported).
> >
> > Patch builds were disabled overnight (for a different reason) but are
> back
> > again.
> >
> > St.Ack
> >
> >
> >
> >
> >
> > On Thu, Aug 20, 2015 at 9:44 AM, Stack <st...@duboce.net> wrote:
> >
> > > On Wed, Aug 19, 2015 at 6:18 PM, Stack <st...@duboce.net> wrote:
> > >
> > >> https://issues.apache.org/jira/browse/BUILDS-109
> > >>
> > >
> > > This became https://issues.apache.org/jira/browse/INFRA-10150 which
> > > becomes BUILDS-111.
> > >
> > > Errant left-over java processes sucking up resources. Trunk is
> > > particularly susceptible because we spin up way more threads than in
> > other
> > > branches (async rpc client). See HBASE-14262 for follow-up.
> > >
> > > St.Ack
> > >
> > >
> > >>
> > >> St.Ack
> > >>
> > >> On Wed, Aug 19, 2015 at 5:38 PM, Stack <st...@duboce.net> wrote:
> > >>
> > >>> On Wed, Aug 19, 2015 at 5:24 PM, Stack <st...@duboce.net> wrote:
> > >>>
> > >>>> On Wed, Aug 19, 2015 at 5:08 PM, Sean Busbey <bus...@cloudera.com>
> > >>>> wrote:
> > >>>>
> > >>>>> I don't recall anything, but I'll go dig through the infra@ list
> > >>>>> later this
> > >>>>> evening to make sure.
> > >>>>>
> > >>>>> Is this on all the H* build hosts?
> > >>>>>
> > >>>>>
> > >>>> Seemingly, yes.
> > >>>>
> > >>>> I can look at infra changes, no prob. Sean. Was just wondering if
> > >>>> anyone knew of a change.
> > >>>>
> > >>>>
> > >>> I don't see anything in INFRA JIRA mailing list related.
> > >>>
> > >>> Let me file an INFRA BUILDS issue and see if anything comes of it.
> > >>>
> > >>> St.Ack
> > >>>
> > >>>
> > >>>
> > >>>> Thanks,
> > >>>> St.Ack
> > >>>>
> > >>>>
> > >>>>
> > >>>>> On Wed, Aug 19, 2015 at 7:05 PM, Stack <st...@duboce.net> wrote:
> > >>>>>
> > >>>>> > Trunk builds are failing with OOME, can't create native thread
> > >>>>> > (HBASE-14262)
> > >>>>> >
> > >>>>> > Anyone know of any recent INFRA changes?
> > >>>>> >
> > >>>>> > Imperfect triage has our problem starting in build #15134, Aug
> 18,
> > >>>>> 2015
> > >>>>> > 5:10:02 AM
> > >>>>> >
> > >>>>> > Otherwise, was going to mess with heap sizings and trying to make
> > >>>>> the big
> > >>>>> > tests smaller so we can get trunk passing again.
> > >>>>> >
> > >>>>> > Thanks,
> > >>>>> > St.Ack
> > >>>>> >
> > >>>>>
> > >>>>>
> > >>>>>
> > >>>>> --
> > >>>>> Sean
> > >>>>>
> > >>>>
> > >>>>
> > >>>
> > >>
> > >
> >
>
>
>
> --
> Sean
>
