Archived community.zenoss.org | full text search
Skip navigation
2414 Views 4 Replies Latest reply: Nov 2, 2011 3:51 PM by Ryan Matte RSS
Dave Bucci Rank: White Belt 21 posts since
Oct 21, 2011
Currently Being Moderated

Oct 28, 2011 4:26 PM

Remodel of device not picking up increased CPU count in VM

Hi, all - we recently added CPUs to a number of VMs within our system.  When we go to the device info, however, in the "Components -> Processors" info the old number of CPUs (4) is displayed, instead of the new # (8).  In fact, I believe that the 8 CPUs are being used for calculations -- because for example, the "Load Average" graph went from a maximum of 8 "processes" (core equivalents) to 16, which makes complete sense as these are allocated as virtual dual core CPUs.  And when I do an snmpwalk against hrProcessorLoad, 8 CPUs are listed.

 

I've verifed that the remodel did occur as expected per the usual 12-hour cycle; I also manually initiated on the device screen (through the gear widget "Model Device" choice), and at the CLI.  None of these have upped the component processor count.

 

At this point, it doesn't seem to matter, since all the statistics being gathered reflect the correct # of CPUs ... but it is troubling, and I am wondering if it will have downstream implications.

 

Any advice?

 

--

dave bucci

  • Matthew Kitchin (public) Rank: Green Belt 170 posts since
    Nov 12, 2009

    I just did the exact same thing except it was more CPUs in a phyiscal servers. I couldn;t find any way to make it pick up the change except delete the device and re-add it. I retained all my perfomance history and events.

  • Matthew Kitchin (public) Rank: Green Belt 170 posts since
    Nov 12, 2009

    When you delete the item, it will pop up a box asking if you also want to delete history and events.

    I only had 2 servers to do, so it wasn't a big deal. It was quicker than spending more time trying to figure out why it wouldn't update. The one other thing I was going to try was to change something else to see if it would force it to remodel everything thing.

  • Ryan Matte ZenossMaster 653 posts since
    Mar 26, 2009

    Just a couple of things to point out.

     

    Firstly, the increase from 8 to 16 in load average has nothing to do with Zenoss.  The load average is a value calculated by the server itself and represents the number of processes either running or waiting on a CPU/core.  Since it now has more cores to offer running processes this value will increase on a busy server.

     

    Secondly, there is a difference between the number of physical (or virtual) CPUs in a system and the number of cores.  hrProcessorLoad represents the number of cores, not physical CPUs.  You could potentially have 1 processor in a system with 4 cores and hrProcessorLoad would show 4 values.  In terms of the number of processors shown in Zenoss, it would show 1 since there is 1 physical processor.

     

    When you are adding virtual CPUs to a server, you are really only adding virtual cores.

     

    If actual physical CPUs are added and it's not updating then you either have the device set as locked, or there is some bug in Zenoss preventing the update from taking place.

     

    If you identify that it is in fact a bug then go to http://dev.zenoss.org/trac, login with zenoss/zenoss and file a bug report.

More Like This

  • Retrieving data ...

Legend

  • Correct Answers - 4 points
  • Helpful Answers - 2 points