none
Reboot the head node RRS feed

  • Question

  • Hey,

         I'm wondering how to reboot the head node when I have jobs running. Will rebooting the head node end all current jobs? I imagine the head node must be up for the jobs to continue and the only way to successfully reboot the head node without loss is to have all jobs end before rebooting the head node.


         Is this the case? Please guide me!

    Thanks,

    Eric

    Thursday, December 9, 2010 8:45 PM

Answers

  • Only the jobs that actually run on the HN will be canceled. All other jobs running on other CN will continue to run and it is usually safe to reboot the HN.

    Depending on how the job was configured and scheduled the tasks that ran on HN might get rescheduled.

    Wednesday, January 12, 2011 11:31 PM

All replies

  • Hi Eric,

    If you reboot the head node, then you are pretty much killing the scheduler service.  Any job that *any* user us running will be cancelled, and any running tasks that the jobs have will be cancelled as well.  

    You can wait for the jobs to finish, or you can just reboot and the running tasks that the jobs have will be cancelled.  Once the HN reboots, you can re-eque your job and any cancelled tasks will start from the beginning (unless the tasks were running some program that can resume where it left off).

    If you have a failover HN set up, then the cluster will continue working when the HN reboots since the failover head node will continue the scheduling w/o interruption of service.

    Hope this helps.

    -Christian

    Thursday, December 9, 2010 10:53 PM
  • Only the jobs that actually run on the HN will be canceled. All other jobs running on other CN will continue to run and it is usually safe to reboot the HN.

    Depending on how the job was configured and scheduled the tasks that ran on HN might get rescheduled.

    Wednesday, January 12, 2011 11:31 PM