1. guilhem
2. Brett
3. Emiliano
4. cloph

* CI
  + current mac bottleneck
    - Brett: can pick the mac mini same day if in stock (guilhem to poke Florian
      next week)
    - cloph will rent a temporary box to fill the gap in the meantime if things
      are too tight — probably a good idea anyway due to high demand on the
  + multijob plugin:
    - cloph: can be upgraded to >v611 now, plugin is deprecated but let's keep
      it for now: the visualization for the the alternative is lacking
    - might have to manually kill jobs when safe-restarting due to a recursion
      bug , cf.
      IRC snippet
        <cloph_away> noel_grandin: sad, happened a few times recently. Played
                     with converting it to pipeline based workflow (since
                     multijob now got deprecated) - but pipleline jobs don't
                     have proper visualization on the individual build's status
                     page :-(
        <noel_grandin> maybe they are in that new "blue ocean" part?
        <noel_grandin> proper visualisation that is
        <cloph_away> noel_grandin: yes, blue ocean has better visualization, but
                     requires moving to a separate "shadow" instance...
        <noel_grandin> oy vey
        <cloph_away> e.g -
                     no listing of individual jobs on the status page
        <cloph_away> and worse on
                     the skipped build (one of the android ones is skipped) is 
shown as success
        <cloph_away> there is visualization on -
                     but of course impossible to find your build in that list...
        <cloph_away> blue ocean shows the android one as skipped, but also
                     requries you to click individual builds to see stats...
    - cloph: pondered replacing it by gerrit checks( ), but that is also
      deprecated now…
      . cloph: can also use separate labels (2 for android, 2 for linux, 1 for
        window, 1 for mac) but that might clutter the page, benefit is that 
        result shows up immediately
      . let's keep things as is for now
  + very long restarts
    - metadata-accumulation workaround job:
      . parsing all historic builds does slow down startup, but doesn't explain
        multi-hours of difference from one restart to another…
    - Noel offered to investigate if handed over threadumps “something like a
      bash loop doing a jstack every 10 seconds for a few minutes” during a long
      restart. `while true; do jattach PID threaddump; sleep 10; done 
  + memory leaks(?)
      WARNING#011hudson.XmlFile#replaceIfNotAtTopLevel: JENKINS-45892: 
reference tgerrit_windows #133164 being saved from unexpected 
      [124056.875s][warning][gc,alloc] Handling GET 
/view/disabled/job/lo_callgrind_linux/12344/logText/progressiveText from … : 
Jetty (winstone)-99792: Retried waiting for GCLocker too often allocating 
4850690 words
      WARNING#011o.e.j.s.h.ContextHandler$Context#log: Error while serving
      java.lang.OutOfMemoryError: Java heap space
    - cloph: disabled the build failure analyzer on the huge callgrind logs,
      let's see if that helps (say in a week time)
      . [edit, guilhem] so far so good
* Next call: Sept 20 a 16:30 UTC


To unsubscribe e-mail to:
Posting guidelines + more:
List archive:
Privacy Policy:

Reply via email to