locked
CPU Load of Head Node increase in Big Steps RRS feed

  • Question

  • Hello Togehter,

    we are running Microsoft HPC 2012 R2. Our Head Node is splitted into two machines, one machine holds the Microsoft SQL Database, and the other Machine has the Microsoft HPC 2012 R2 Application installed.
    The Problem i describe here is only occuring on the Applikation Server (this one with Microsoft HPC installed.)
    I observed the behaviour, that the CPU Load of our Application Head node is increasing in some steps.
    In the past, when our Head Node had not so much CPU Power (2 CPU System), he reached very fast the 100% CPU Load in several hours.
    Now we updated our Head Node with a 8 Core System, and the CPU Load is still increasing very fast.
    But it does never reach the 100 % which is very good. But still the behaviour i don't understand, 
    because in a graph with the cpu load over time, we have following behaviour. Please see following Graph:


    So my question is, if this behaviour of a Head Node is normal, and if somebody could explain why the CPU Load of the HeadNode goes this way,
    because every time we reboot, everythings works fine, with a much lower cpu load.
    We have some other Issues like not assigning enough resources to a job, comunication problems to nodes.
    slow reaction time of our server, where we don't know, if this is based on this cpu load problem.
    The only thing we notice is that when we have such kind or problems, it helps to reboot our head node......

    Any information would be welcome and helpful, to understand the whole thing a little bit better.

    Thank you very much in advance,

    Bobby
    • Edited by Bobby013 Friday, February 27, 2015 7:16 AM Added Image
    Thursday, February 26, 2015 9:02 AM