Re: [ovirt-devel] Moving F21 slaves offline / disabling failing F21 jenkins jobs

I disagree. Let me explain what the purpose of CI is, in case it got lost in the noise of the recent failures. CI is a validation system for PRODUCTION code (at least the jobs that test merged commits), It is not a DEBUGGING system where you test if you code works or not and try out new untested systems, Its purpose is to detect regressions in the code and alert developers when it happens. The normal flow of CI is to be stable all the time and only fail on new regressions, of course there are infra issues we try to solve all the time, so we don't need extra failures due to new untested system. Recently unfortunately it has become more a testing/debugging system rather than production, I think this should stop, since developers are loosing faith in the CI and will eventually ignore the failure (some probably are doing that already). And now to the point in hand - fedora 21. I believe that a test should run on a new system only after it run successfully locally (or on a testing jenkins instance), and approved by the maintainer it works - so i expect each pkg maintainer to test his code / packaging on f21 before deploying it to the production jenkins. Infra team can help with providing test VMs if needed or help in running the code also, but until the code isn't fixed for running on the new OS locally, it shouldn't go on jenkins. The infra team has very little resources in terms of people managing jenkins, and i believe that main focus should be on making sure we don't break existing supported systems (like EL6/EL7/FC20) rather than chasing developers to fix fc21 issues and on the way making a lot of noise on the CI. Once a project has passed localy sanity for a new system, we'd be happy to enable it on CI. I hope this clarifies the scope and purpose of jenkins in oVirt. Eyal. ----- Original Message -----