2 Replies Latest reply on Nov 16, 2017 3:30 PM by rschroeder

    Do your Cisco switches & routers show unexpected high CPU in Orion, that's not explained by a known hardware chip flaw?

    rschroeder

      I built a new custom report to show the Top Cisco Nodes in my network by CPU utilization, and shared it with my Network Team.

       

      One of them did some digging on nodes with unexpectedly high CPU utilization and came up with a Bug Tool entry that suggests any high CPU alarms MIGHT be false alarms. The information offers alternate methods of polling to compensate and correct for false high CPU alerts.

       

      I thought some of you might be interested:

       

       

       

       

       

      Linux kernel cache is leading to confusion over memory usage

      CSCuc40262

       

      Description

       

      Symptom:
      Over time on an IOS-XE Routers, such as ASR1xxx's, ASR9xx, and ISR4xxx routers, and other running IOS-XE code, the kernel will use free space to grow it's cache which will cause the free memory to slowly shrink over time. This can look like the router is experiencing a memory leak and is potentially in danger of running out of memory, when in fact the memory usage is fine since the cache can be freed if the memory is needed.

      This can also cause a false alarm in network management software if the incorrect free memory OID is polled.

      Conditions:
      This is not a hardware dependency so this will happen on any router running IOS-XE over time or upon boot up for low end routers such as ISR43xx routers and others.

      Workaround:
      Instead of looking at the used and free values in "show platform software status control-processor brief" to determine if the memory usage is healthy, instead use the committed memory value. If the committed memory is over 90% and you meet the minimum required DRAM (as specified in Cisco.com), please work with TAC for further investigation. If it is less than 90%, then the memory usage is fine.

      If this issue is causing an alarm in network management software, the following OID should be used instead:

      http://tools.cisco.com/Support/SNMP/do/BrowseOID.do?local=en&translate=Translate&objectInput=1.3.6.1.4.1.9.9.109.1.1.1.1.27

      Further Problem Description:
      The fix for this will make the output of "show platform software status control-processor brief" more clear about how much memory is actually available for use. The Used Memory is accurate after used memory utilization actually cross 88% [warning level]. In other words, this fix is only valid for situations where the memory has reached/crossed the 88% threshold.


      NOTE: this code enhancement is only available in 3.17S release. Any older code or 16.3.1 onwards won't have this code enhancement.