3 Replies Latest reply on Aug 22, 2016 7:35 PM by cgregors

    CPU Utilization is not displaying correctly (Ciso ISR4451)

    jlim13

      Hi,

       

      I'm hoping somebody can help me with this, I've been wondering why my Cisco ISR4451 (Cisco IOS Software, ISR Software (X86_64_LINUX_IOSD-UNIVERSALK9-M), Version 15.5(3)S2, RELEASE SOFTWARE (fc2)) is not showing the correct CPU utilization in NPM.  Yesterday, there was an instance wherein the CPU reached 100% but in Orion, it shows only 8-9% peak contrary to what the actual router is saying, for some reason the NHRP process on this router burst maxing the CPU so I needed to reset the process, but I never got an alert in Orion and when I checked in NPM the utilization did not peak to 100% that's why it did not alert. I'm already using the custom device poller on this post: Cisco ASR Showing 95% 98% Memory Used.

       

        • Re: CPU Utilization is not displaying correctly (Ciso ISR4451)
          jlim13

          Anybody who would like to take stab on this? Appreciate any assistance, suggestion or recommendation the community can share.

           

          Regards!

          • Re: CPU Utilization is not displaying correctly (Ciso ISR4451)
            amarnath_r

            We have this issue in 11.5 and after upgrading to 12 it seems to be fine now.

            • Re: CPU Utilization is not displaying correctly (Ciso ISR4451)
              cgregors

              This is a common problem but if you dissect it a little it will make sense

              1. The Cisco graph which is derived from on-board CPU load history shows the spike
              2. The Orion graph which is derived from SNMP gathered history shows a data gap.
              3. When the Cisco device goes to 100% CPU, it's too busy to respond to SNMP queries from Orion and thus there is no data point to graph
              4. When the CPU comes out of 100% utilization, SNMP can reply to Orion and you get data points again

               

              Observation:  The rising and falling edges of the Orion graph should line up with the device going into and coming out of 100% CPU utilization.

              Observation:  This is not Orion or Cisco version dependent. 

               

              Solution 1:  Increase the SNMP timeout for polling this device (All devices would be a bad idea).  This MAY give SNMP time to reply to Orion.

              Solution 2:  Write an alert that triggers when a device returning valid data points.  This is the UNKNOWN state.

               

              Chris

              1 of 1 people found this helpful