2 Replies Latest reply on Dec 6, 2017 8:42 AM by peter79

    Memory Usage not reporting Properly

    peter79 Enthusiast



      I need some advice.  I currently support VMware environments for multiple customers and recently I've started getting vROPS cases (I have little experience and no training in this area).


      A customer is reporting is seeing deviations between memory usage reported in the OS and the vROPs (sometimes reporting about 10% higher in vROPS).

      This is only happening to 1 VM.

      There are no errors are alerts in vROPs.

      The collection status on the VM is green.

      According to vROPS guest tools are running and current.


      Based on the above info I dont think this is a VMware issue it sounds more like a graph config/interpretation issue. 

      I tried to compare live memory usage on the VM to vROPS however because I'm new to the product I'm not sure I was looking at the graphs correctly.  Additional tech info


      vROPS version 6.4.0 (build 4635874)

      VM OS  Red Hat Enterprise Linux 7 (64 bit).


      Any pointers tips would be greatly appreciated. 

        • 1. Re: Memory Usage not reporting Properly
          sxnxr Expert

          You cant compare the memory reported in vrops to what is being reported in the OS.


          The vrops memory stats does not take into consideration the application memory usage in the VM. SQL for instance will cache store procedures in memory for faster recall but that memory is not active on the hosts. The memory stat in vrops will only show you what is active on the host so you will in most cases see the vrops start much lower than the OS stats.


          vrops did introduce OOTB in guest memory stats without the need for an agent. Not sure if 6.4 supports the in guest and it is also ESXi version and tools version specific someone else here will hopefully know and post the requirements

          1 person found this helpful
          • 2. Re: Memory Usage not reporting Properly
            peter79 Enthusiast

            Actually the app in question is an Oracle DB and the customer did mention about having some of the allocated mem in some form of cache/reserve.  The only problem here is that apparently there is a second DB server which doesn't have the same "reporting issue"