Hi all. Summing up what we learned here plus some of Davor's and my
investigations.

Issue 1: MapR Hadoop
Diagnosis: Use of Maven global repository may have caused test failure due
to caching of MapR's version of Hadoop.
Mitigation: Switched to private Maven repository.
Next Steps: Determine whether private Maven repository fixes problem
permanently or is just a band-aid. This could be a slave-dependent problem.

Issue 2: Jenkins status fails to update.
Diagnosis: Re-running failed pull request builds instead builds the master
branch of the repository. This is due to the Jenkins job configuration.
Mitigation: Submitted new changes to pull request, kicking off a fresh
Jenkins build via webhooks.
Next Steps: Determine how to configure the rebuild job action to actually
rebuild the pull request in question and update the Github status.

Let me know if there's anything incorrect or any information I may have
missed.

Thanks,

Jason

On Mon, Mar 14, 2016 at 12:19 PM, Amit Sela <[email protected]> wrote:

> Jenkins seems satisfied now :)
>
> I guess "Use private Maven repository" property did the job ?
>
> Thanks,
>
> Amit
>
> On Mon, Mar 14, 2016 at 7:27 PM Jason Kuster
> <[email protected]>
> wrote:
>
> > Hi all, Davor asked me to investigate this issue.
> >
> > It seems like Jenkins might see no code changes between the earlier build
> > and the newer ones and thus just reuse old results? Amit, have you tried
> > just changing something trivial in the PR, generating a new commit, and
> > seeing if that kicks off a new Jenkins build?
> >
> > Jason
> >
> > On Mon, Mar 14, 2016 at 10:05 AM, Jean-Baptiste Onofré <[email protected]>
> > wrote:
> >
> > > Let's get in touch to see what's going on.
> > >
> > > Thanks,
> > > Regards
> > > JB
> > >
> > >
> > > On 03/14/2016 06:01 PM, Davor Bonaci wrote:
> > >
> > >> We are investigating.
> > >>
> > >> On Sun, Mar 13, 2016 at 11:49 PM, Amit Sela <[email protected]>
> > wrote:
> > >>
> > >> Thanks JB,
> > >>>
> > >>> I've tried to rebuild (#64) but it seems as if it ignores the PR code
> > >>> although it states it's a rebuild of the original PR build (#61).
> > >>>
> > >>> Was anything else changed ?
> > >>>
> > >>> On Mon, Mar 14, 2016 at 2:48 AM Jean-Baptiste Onofré <
> [email protected]>
> > >>> wrote:
> > >>>
> > >>> Hi Amit,
> > >>>>
> > >>>> I updated the two Beam jobs to use private Maven repository. Let me
> > know
> > >>>> if it helps (else I will revert the change).
> > >>>>
> > >>>> Regards
> > >>>> JB
> > >>>>
> > >>>> On 03/13/2016 07:47 PM, Amit Sela wrote:
> > >>>>
> > >>>>> I don't think I have permissions...
> > >>>>> @JB @davor any of you have permissions to configure Jenkins jobs ?
> > >>>>>
> > >>>>> On Sun, Mar 13, 2016 at 4:29 PM Andreas Veithen <
> > >>>>>
> > >>>> [email protected]>
> > >>>>
> > >>>>> wrote:
> > >>>>>
> > >>>>> You should probably enable the "Use private Maven repository"
> option
> > >>>>>>
> > >>>>> in
> > >>>
> > >>>> the
> > >>>>
> > >>>>> Jenkins build.
> > >>>>>>
> > >>>>>> Andreas
> > >>>>>>
> > >>>>>> On Sun, Mar 13, 2016 at 8:11 AM, Amit Sela <[email protected]>
> > >>>>>>
> > >>>>> wrote:
> > >>>>
> > >>>>>
> > >>>>>> I don't see anything either, but I suspect it's some sort of a
> cache
> > >>>>>>>
> > >>>>>> in
> > >>>
> > >>>> Jenkins maven.
> > >>>>>>>
> > >>>>>>> When I tried to build the original spark-dataflow project
> > (Cloudera)
> > >>>>>>> against Beam instead of Dataflow, I removed all Cloudera
> > repositories
> > >>>>>>>
> > >>>>>> from
> > >>>>>>
> > >>>>>>> the pom and even removed the Cloudera hadoop artifacts from my
> > >>>>>>> .m2/repository but it kept downloading hadoop CDH... Only after I
> > >>>>>>>
> > >>>>>> deleted
> > >>>>
> > >>>>> my entire ~/.m2/repository it downloaded Apache Hadoop.
> > >>>>>>>
> > >>>>>>> I'm not suggesting we delete Jenkins' .m2/repository (yet...)
> but I
> > >>>>>>>
> > >>>>>> think
> > >>>>
> > >>>>> it's some sort of cache and I'm not sure how to resolve it.
> > >>>>>>>
> > >>>>>>> You gotta admit it's strange that it downloads MapR hadoop
> > >>>>>>>
> > >>>>>> artifacts....
> > >>>>
> > >>>>>
> > >>>>>>> On Sun, Mar 13, 2016 at 9:18 AM Davor Bonaci
> > >>>>>>>
> > >>>>>> <[email protected]
> > >>>
> > >>>>
> > >>>>> wrote:
> > >>>>>>>
> > >>>>>>> I've added "help:effective-settings" goal to the pre-commit
> Jenkins
> > >>>>>>>> project, and rebuilt your pull request in Jenkins #62.
> > >>>>>>>>
> > >>>>>>>> I didn't see anything suspicious; see console output with
> > effective
> > >>>>>>>> settings.xml here [1].
> > >>>>>>>>
> > >>>>>>>> [1] https://builds.apache.org/job/beam_PreCommit/62/console
> > >>>>>>>>
> > >>>>>>>> On Sat, Mar 12, 2016 at 10:45 PM, Amit Sela <
> [email protected]
> > >
> > >>>>>>>>
> > >>>>>>> wrote:
> > >>>>>>>
> > >>>>>>>>
> > >>>>>>>> OS=looks. Sorry, phone autocorrect ;)
> > >>>>>>>>>
> > >>>>>>>>> On Sun, Mar 13, 2016, 08:39 Amit Sela <[email protected]>
> > >>>>>>>>>
> > >>>>>>>> wrote:
> > >>>
> > >>>>
> > >>>>>>>>> Well, I can't run mvn dependency:tree but the downloading of
> MapR
> > >>>>>>>>>>
> > >>>>>>>>> Hadoop
> > >>>>>>>>
> > >>>>>>>>> dependencies OS suspicious... I remember I had this issue once
> > >>>>>>>>>>
> > >>>>>>>>> with
> > >>>
> > >>>> Cloudera distribution and I resolved it by deleting my entire
> > >>>>>>>>>>
> > >>>>>>>>> .m2/repository
> > >>>>>>>>>
> > >>>>>>>>>>
> > >>>>>>>>>> And Travis runs the same tests successfully...
> > >>>>>>>>>>
> > >>>>>>>>>> On Sun, Mar 13, 2016, 08:35 Davor Bonaci
> > <[email protected]
> > >>>>>>>>>>
> > >>>>>>>>>
> > >>>>>>> wrote:
> > >>>>>>>>>
> > >>>>>>>>>>
> > >>>>>>>>>> Jenkins jobs currently use "default Maven settings" and
> "default
> > >>>>>>>>>>>
> > >>>>>>>>>> Maven
> > >>>>>>>
> > >>>>>>>> global settings". We have some (limited) ability to view and
> > >>>>>>>>>>>
> > >>>>>>>>>> change
> > >>>>>>
> > >>>>>>> these
> > >>>>>>>>>
> > >>>>>>>>>> settings ourselves.
> > >>>>>>>>>>>
> > >>>>>>>>>>> Can we definitely confirm this is the root-cause?
> > >>>>>>>>>>>
> > >>>>>>>>>>> On Sat, Mar 12, 2016 at 8:49 AM, Amit Sela <
> > [email protected]
> > >>>>>>>>>>>
> > >>>>>>>>>>
> > >>>> wrote:
> > >>>>>>>>>
> > >>>>>>>>>>
> > >>>>>>>>>>> I've issued PR
> > https://github.com/apache/incubator-beam/pull/42
> > >>>>>>>>>>>>
> > >>>>>>>>>>> and
> > >>>>>>>
> > >>>>>>>> while
> > >>>>>>>>>>>
> > >>>>>>>>>>>> Travis seems happy about it, Jenkis is not.
> > >>>>>>>>>>>>
> > >>>>>>>>>>>> Checking the build console
> > >>>>>>>>>>>> https://builds.apache.org/job/beam_PreCommit/61/console I
> see
> > >>>>>>>>>>>>
> > >>>>>>>>>>> that
> > >>>>>>>
> > >>>>>>>> for
> > >>>>>>>>>
> > >>>>>>>>>> some
> > >>>>>>>>>>>> reason it downloads MapR hadoop artifacts such as:
> > >>>>>>>>>>>>
> > >>>>>>>>>>>> [INFO] Downloading:
> > >>>>>>>>>>>>
> > >>>>>>>>>>>>
> > >>>>>>>>>>>>
> > >>>>>>>>>>>
> > >>>>>>>>>
> > >>>>>>>>
> > >>>>>>>
> > >>>>>>
> > >>>>
> > >>>
> >
> https://repository.jboss.org/nexus/content/repositories/releases/org/apache/hadoop/hadoop-client/2.7.0-mapr-1509/hadoop-client-2.7.0-mapr-1509.jar
> > >>>
> > >>>>
> > >>>>>>>>>>>> instead of Apache Hadoop from Maven Central.
> > >>>>>>>>>>>>
> > >>>>>>>>>>>>
> > >>>>>>>>>>>> It seems that the issues causing the tests to fail are
> related
> > >>>>>>>>>>>>
> > >>>>>>>>>>> to
> > >>>>>>
> > >>>>>>> the
> > >>>>>>>>
> > >>>>>>>>> MapR distribution..
> > >>>>>>>>>>>>
> > >>>>>>>>>>>>
> > >>>>>>>>>>>>
> > >>>>>>>>>>>> Any reason for that ? How/Where does Jenkins configure it's
> > >>>>>>>>>>>>
> > >>>>>>>>>>> settings.xml ?
> > >>>>>>>>>>>
> > >>>>>>>>>>>>
> > >>>>>>>>>>>>
> > >>>>>>>>>>>> Thanks,
> > >>>>>>>>>>>>
> > >>>>>>>>>>>>
> > >>>>>>>>>>>> Amit
> > >>>>>>>>>>>>
> > >>>>>>>>>>>>
> > >>>>>>>>>>>
> > >>>>>>>>>>
> > >>>>>>>>>
> > >>>>>>>>
> > >>>>>>>
> > >>>>>>
> > >>>>>
> > >>>> --
> > >>>> Jean-Baptiste Onofré
> > >>>> [email protected]
> > >>>> http://blog.nanthrax.net
> > >>>> Talend - http://www.talend.com
> > >>>>
> > >>>>
> > >>>
> > >>
> > > --
> > > Jean-Baptiste Onofré
> > > [email protected]
> > > http://blog.nanthrax.net
> > > Talend - http://www.talend.com
> > >
> >
> >
> >
> > --
> > -------
> > Jason Kuster
> > Cloud Dataflow/Apache Beam (incubating)
> >
>



-- 
-------
Jason Kuster
SETI - Cloud Dataflow

Reply via email to