Mathieu Cantin added a comment - 2015-03-12 14:47 You have the option "Kill the phase on:" for each sub-jobs on a phase. When this is on "Failure", this kill all others builds when this sub-job has failed. Select "never" to never kill others sub-jobs. The option "Abort all other job" is when the job is aborted manually by an user.

Simon Weber added a comment - 2015-03-12 14:52 I'm not sure that has anything to do with the problem? The problem is that multijob will sometimes lose track of a child job and wait forever for its result. That forces us to abort the parent build (or else the parent job would run indefinitely).

Mathieu Cantin added a comment - 2015-03-12 15:45 Sorry, I see the problem. Do you have the "Execute concurrent builds if necessary" checked on child_job ?

Simon Weber added a comment - 2015-03-12 16:03 > Sorry, I see the problem Oh, no worries; I could have done a better job describing the problem. > Do you have the "Execute concurrent builds if necessary" checked on child_job? Yes; on both child_job and parent_job. I had originally suspected this only happened when job numbers were interleaved due to another concurrent build, but that didn't happen in an example I just triggered now. I'll check the logs to see if there's anything interesting.

Simon Weber added a comment - 2015-03-12 16:34 The logs aren't too informative. My comments are in [ square brackets ]. parent_job log Starting build job child_job. Starting build job child_job. Finished Build : #6631 of Job : child_job with status : SUCCESS Build timed out (after 20 minutes). Marking the build as aborted. [ this is from a plugin we use to timeout automatically ] Aborting all subjobs. Finished Build : #6630 of Job : child_job with status : ABORTED [ note that the parent job knows about the correct subjob since it knows which to abort ] Build was aborted Finished: ABORTED Jenkins log: Mar 12, 2015 2:34:40 PM INFO hudson.model.Run execute child_job #6631 main build action completed: SUCCESS Mar 12, 2015 2:34:45 PM INFO hudson.model.Run execute child_job #6630 main build action completed: SUCCESS [ this subjob did finish successfully and on time, but the parent job missed its result ] <snip> Mar 12, 2015 2:36:52 PM SEVERE hudson.model.Executor run [I don 't think this is relevant - it happens later than the job finishes - but I figured I' d include it ] Executor threw an exception java.util.NoSuchElementException at jenkins.model.lazy.LazyLoadRunMapEntrySet$1.next(LazyLoadRunMapEntrySet.java:76) at jenkins.model.lazy.LazyLoadRunMapEntrySet$1.next(LazyLoadRunMapEntrySet.java:63) at java.util.AbstractMap$2$1.next(AbstractMap.java:385) at hudson.util.RunList.subList(RunList.java:137) at hudson.tasks.LogRotator.perform(LogRotator.java:124) at hudson.model.Job.logRotate(Job.java:449) at hudson.model.Run.execute(Run.java:1823) at hudson.model.FreeStyleBuild.run(FreeStyleBuild.java:43) at hudson.model.ResourceController.execute(ResourceController.java:89) at hudson.model.Executor.run(Executor.java:240)

Mathieu Cantin added a comment - 2015-03-16 19:53 Good new (or bad). This seen a change on the API of Jenkins 1.596.1 (we have updated last friday). I work on a fix.

Simon Weber added a comment - 2015-03-16 20:36 Is the api change you're thinking of in that specific version? We're still on version 1.593, and I think we saw this behavior on a version < 1.565.1 as well.

Daniel Beck added a comment - 2015-03-16 20:40 If something hangs, thread dumps would be helpful. https://wiki.jenkins-ci.org/display/JENKINS/Obtaining+a+thread+dump