Alarms for high CPU with Host_SystemEdge indicate an individual CPU issue when in reality the alarm is for the aggregate of all CPU

Products

• CA Spectrum

Issue/Introduction

CPU monitoring in CA Spectrum for Host_SystemEdge may cause confusion as CA Spectrum monitors the Entity Mib which has an aggregate value of CPU.  The event/alarm indicate a problem noting individual CPU (Event 0x10f03).  Here is an example:

Major

Jun 1, 2016 11:09:14 AM EDT

Sim6850

High CPU Utilization. The utilization of 13% for CPU instance 1 named 'CPU: 1' has exceeded the 5% threshold on model Sim6850 for more than the acceptable time period.

System

0x10f03

10

 

Environment

Release: Any version of Spectrum
Component:

Cause

The threshold is looking at a table (NRM_DeviceCPUUtilization) that for SystemEDGE is only populated by cpuTotalUtilPercent which represents the aggregate (average) CPU utilization across all CPU instances on the host.  For other devices (rfc2790, for example), the NRM_DeviceCPUUtilization table has entries for individual CPU instances. 

As can be seen by the mib description, cpuTotalUtilPercent is an aggregate of all CPU:

cpuTotalUtilPercent OBJECT-TYPE

SYNTAX INTEGER (-2147483648 .. 2147483647)

ACCESS read-only

STATUS mandatory

DESCRIPTION

"The percentage of time (over the sample period) that the system's CPUs were utilized."

::= { cpuGroup 13 }

-- 1.3.6.1.4.1.546.13.13

Resolution

There are different ways to handle this, however here are two suggested solutions:

1.  You can use the Event Configuration gui to edit Event 0x10f30 to update the text in the event noting that for Host_SystemEdge this may be an aggregate value.

2.  Create (or edit) the <SPECROOT>/custom/Events/Ctron_Gen_HOST/Host_sysetmEDGE/EventDisp and add the following which will change the event generated on the Host_SystemEDGE model to note that the alarm is for the aggregate CPU:

0x00010f01 S-   R Aprisma.EventPairTimeAttr, 0x00010f02, "0x00010f09 -:-", 0x12bce

0x00010f09 E 50 A 2, 0x00010f09

0x00010f02 E 50 C    0x00010f09

You will then need to press the "Update Event Configuration" button on the VNM in the Component Detail -> SpectroSERVER Control area.

Notes:

CA Spectrum utilizes data points for the time duration that is specified to determine if an alarm should be generated or not.  For example, if you set the Duration for 30 seconds, Spectrum does not poll every second for the value and take an average.  Spectrum will poll at the 0 and 30 second time point.  If both data points exceed the threshold configured, an event/alarm is generated.  

The polling interval of CA Spectrum may not be the same time as any other application (even other CA products) as CA Spectrum does not synchronize polling intervals for data points with any other application.