locked
How to determine if running on a MS HPC headnode RRS feed

  • Question

  • Hi,

    I am developing a Python tool for submitting jobs (reservoir simulation case) to Microsoft Compute Cluster (CCS1 and HPC) or running directly on a Windows machine.
    The tool runs fine when submitting jobs with 'job submit' command, however user is also able to run an executable directly on a Windows machine and here the problem starts - I need to check if I am running on a headnode to pick up different executable (the one which uses MS MPI rather than MPI).

    In the past I was checkng for presence of CcsScheduler process running, but it doesn't seem to be the best approach since HPC renamed the process and my condition is no longer valid.

    I will appreciate any help.

    Thanks,
    Krzysztof
    Thursday, May 28, 2009 2:44 PM

Answers

  • I believe the v2 scheduler process is HPCScheduler if that helps at all.  You can also try running cluscfg view (or from PSH Get-HpcClusterOverview) which will provide you the name of the scheduler you are connecting too . . . you can just compare that to the name of the machine you're currently running on.

    Thanks,
    Josh
    -Josh
    Thursday, May 28, 2009 7:10 PM
    Moderator
  • Yep, we use the name of the head node to identify the cluster.
    Wednesday, June 24, 2009 10:06 PM
    Moderator

All replies

  • I believe the v2 scheduler process is HPCScheduler if that helps at all.  You can also try running cluscfg view (or from PSH Get-HpcClusterOverview) which will provide you the name of the scheduler you are connecting too . . . you can just compare that to the name of the machine you're currently running on.

    Thanks,
    Josh
    -Josh
    Thursday, May 28, 2009 7:10 PM
    Moderator
  • Hi Josh,

    Thanks for reply.

    Another question emerges: 'is Cluster Name always the same as headnode machine's name?'.
    I have seen quite a few CCS/HPC clusters and in every case the cluster's and headnode's name were the same, however is it always the case?

    Regards,
    Krzysztof
    Friday, May 29, 2009 8:48 AM
  • Yep, we use the name of the head node to identify the cluster.
    Wednesday, June 24, 2009 10:06 PM
    Moderator