How to monitor CPU processes instead of CPU as a whole

[joch2joch]

Hi,

We have been using SNMP tools to monitor our router CPU utilization. While it is helpful to see the trend/peaks....we really need to be able to see the exact CPU process that is contributing to the CPU peaks.

Does anyone know how we can achieve this?

Thanks for the wisdom!

Joe

[beth-martin]

You can try using the suite of SNMP tools such as UCD's snmpwalk utility.This is available for free

download at http://sourceforge.net/project/showfiles.php?group_id=12694.You could also refer to the webpage which explains how the CPU utilization can be collected using SNMP.

The URL is http://www.cisco.com/warp/public/477/SNMP/collect_cpu_util_snmp.html

[ioiioi2000]

Thers is a problem with using snmp to monitor CPU's unitilizatioin.

The OID which I use in mrtg is 1.3.6.1.4.1.9.9.109.1.1.1.1.5.1, but the graphic is not as accurate as the result of "show process cpu history".

for example:

show process cpu history at 16:30

100 1225221114111141112141112421212322124112214111131122231111

90

80

70

60

50

40

30

20

10 *

0....5....1....1....2....2....3....3....4....4....5....5....

0 5 0 5 0 5 0 5 0 5

CPU% per minute (last 60 minutes)

* = maximum CPU% # = average CPU%

data come from prtg which is similer to mrtg:

2005-4-4 16:25 - 16:30 1.000

2005-4-4 16:20 - 16:25 1.000

2005-4-4 16:15 - 16:20 1.000

2005-4-4 16:10 - 16:15 1.000

2005-4-4 16:05 - 16:10 1.000

2005-4-4 16:00 - 16:05 1.000

2005-4-4 15:55 - 16:00 1.000

2005-4-4 15:50 - 15:55 1.000

2005-4-4 15:45 - 15:50 1.000

2005-4-4 15:40 - 15:45 0.900

2005-4-4 15:35 - 15:40 1.000

2005-4-4 15:30 - 15:35 0.900

2005-4-4 15:25 - 15:30 1.000

2005-4-4 15:20 - 15:25 1.000

2005-4-4 15:15 - 15:20 1.000

2005-4-4 15:10 - 15:15 1.000

2005-4-4 15:05 - 15:10 1.000

2005-4-4 15:00 - 15:05 1.000

2005-4-4 14:55 - 15:00 1.000

so, which one I should trust?

[joch2joch]

Hmm, let me repeat my objective here: I don't want to view the CPU as a whole.

I want to monitor EACH CPU process over a period of time

Does anyone know if this is possible? Are there any Cisco Systems specialist monitoring this thread?

Thanks!

[lrian]

CISCO-PROCESS-MIB is probably what you're looking for - it shows the cpu utilization, by process, for each cpu

Note that there can be multiple CPUs in the box. For example, a cat6500 in native mode shows a switch processor (probably the L2 supervisor card) and a route processor (L3 MSFC)

snmpwalk cat6000 cpmCPUTotalPhysicalIndex

CISCO-PROCESS-MIB::cpmCPUTotalPhysicalIndex.1 = INTEGER: 5015

CISCO-PROCESS-MIB::cpmCPUTotalPhysicalIndex.2 = INTEGER: 5001

snmpwalk cat6000 entPhysicalDescr |grep CPU

ENTITY-MIB::entPhysicalDescr.5001 = STRING: CPU of Switching Processor

ENTITY-MIB::entPhysicalDescr.5015 = STRING: CPU of Routing Processor

no idea how to figure out which cpu is which on a 7500. The RSP counts as a CPU along with any VIPs but, at least on our 7500s, cpmCPUTotalPhysicalIndex is always zero :-(

snmpwalk c7500 cpmCPUTotalPhysicalIndex

CISCO-PROCESS-MIB::cpmCPUTotalPhysicalIndex.1 = INTEGER: 0

CISCO-PROCESS-MIB::cpmCPUTotalPhysicalIndex.2 = INTEGER: 0

CISCO-PROCESS-MIB::cpmCPUTotalPhysicalIndex.3 = INTEGER: 0