Uploaded image for project: 'Hadoop Map/Reduce'
  1. Hadoop Map/Reduce
  2. MAPREDUCE-5142

MR AM unregisters with state KILLED when an error causes dispatcher to shutdown

    XMLWordPrintableJSON

Details

    • Bug
    • Status: Open
    • Major
    • Resolution: Unresolved
    • 2.0.3-alpha, 0.23.5
    • None
    • None
    • None

    Description

      RMCommunicator sets final state to KILLED if the job is in a running state and isSignalled is set to true.

            } else if (jobImpl.getInternalState() == JobStateInternal.KILLED
                || (jobImpl.getInternalState() == JobStateInternal.RUNNING && isSignalled)) {
              finishState = FinalApplicationStatus.KILLED;
            } else if (jobImpl.getInternalState() == JobStateInternal.FAILED
                || jobImpl.getInternalState() == JobStateInternal.ERROR) {
              finishState = FinalApplicationStatus.FAILED;
      

      This happens when any uncaught exception in any event handler ends up causing the AsyncDispatcher to trigger a shutdown. In such a scenario, even though the AM actually failed due to some error, its actual state ends up as KILLED.

      Attachments

        Activity

          People

            Unassigned Unassigned
            hitesh Hitesh Shah
            Votes:
            0 Vote for this issue
            Watchers:
            5 Start watching this issue

            Dates

              Created:
              Updated: