Hi, sorry for late response -- i am approaching the usual pre-christmas stackoverflow mode .
Actually the ci job is right now more or less a proof of concept, since starting of docker containers shows massive race conditions. I recommend to define a pre-job of to run docker run -d --name=nexus sonatype/nexus on each jenkins slave before starting Bigtop-trunk-packages Not to forget to remove the if docker logs .. fi from the Bigtop-trunk-packages job. And a post-job on each of the jenkins slaves. docker stop nexus docker rm -v nexus (Important to use -v with docker-rm !) That way we start with new nexus containers each time. I haven't done such a thing before (How do you force to run a job on each slave?) Evans, do you know how to configure such a thing? Olaf > Am 08.12.2015 um 09:00 schrieb Konstantin Boudnik <[email protected]>: > > This is actually super-cool! Thanks for fixing this thingy. You totally rock! > > From looking at the way we run the nexus container doesn't seem to to retain > the M2 repository cache, does it? I am wondering if by keeping the cache > in-between the runs we can make the builds even more autonomous. > > Cos > > P.S. I wish I knew anything about Hue to help with it. > > On Tue, Dec 08, 2015 at 08:36AM, Olaf Flebbe wrote: >> Hi, >> >> Seems like the caching of artifacts almost the does an Allstar-Blues! >> http://ci.bigtop.apache.org/view/Packages/job/Bigtop-trunk-packages/69/. The >> failures because of network issues are (almost) gone. >> >> I observed the only failure >> http://ci.bigtop.apache.org/view/Packages/job/Bigtop-trunk-packages/69/BUILD_ENVIRONMENTS=opensuse-13.2,COMPONENTS=hue,label=docker-slave/ >> several time at my site, too. It is a little bit funny since it should only >> detect the presence of a preinstalled header file and this occasionally >> fails on all linux distros at random times. Maybe this is a docker issue. >> If someone has an idea, please chime in. >> >> >> Olaf >> > >
signature.asc
Description: Message signed with OpenPGP using GPGMail
