none
HPC Cluster Manager (2012 R2 Update 1) does not show Monitoring Data of Clients.

    Question

  • Hello Together,

    In the Cluster Manager we don't see the monitoring data of most workstation nodes.

    The clients are only marked with a "X".

    When i check the event log insside one client, i find following Error Message regulary:

        Failed to add counter to query. System.ComponentModel.Win32Exception (0x80004005): PDH_AddCounterFailed
        at Microsoft.Hpc.Monitoring.MetricCollector.AddCounter(String path, Int32 metricId, Int32 instanceId)

    I tried to reboot the node, which had this errors, but this didn't helped.

    What can we do, to get the HPC Monitoring working?

    Thank you very much in advance,

    best regards,

    Bobby

    Friday, August 14, 2015 1:53 PM

All replies

  • Hi, Bobby,

    from the error, most performance counter cannot be added, from the error code, it means cannot find those counters,

    So can you do the following on one of the error nodes, such as LUD0102G,

    Open Performance Monitor, try add the above failed counters, to see whether the counter exist, and can get the value,

    For example, \Processor(_Total)\% Processor Time means Counter category is “Processor”, Counter Name is “% Processor Time”, instance is “_Total”

    Or you can also

    open cmd console (run as administrator),

    Run the following command can check the system performance counter status

    lodctr /Q

    Monday, August 17, 2015 8:05 AM
  • error 0x80004005, is a system general, refer to http://www.0x80004005.net/
    Monday, August 17, 2015 8:06 AM
  • Hi Yongjun,

    thank you very much for your response. I tried to add a Performance Counter as you described with the Performance Monitor, and this worked well.

    In the mean time we found out, that the node 'LUD3302M' has a different Windows Image installed, and maybe this is the reason why it is working there, and at the other not.

    Do you know some reasons like policy settings, firewall,.... which can also result in this behaviour ?

    best regards,

    Bobby



    • Edited by Bobby013 Tuesday, August 18, 2015 2:41 PM typing error
    Tuesday, August 18, 2015 2:41 PM
  • Hi, Bobby,

    The error occurred during AddCounter in HpcMonitoringClient service, until now, it doesn't need the network with head node, and for firewall policy, during install HPC, we should already apply firewall rules, (for monitoring, HpcMonitoringClient use UDP port 9894 to talk with head node).

    Wednesday, August 19, 2015 1:32 AM