NPM Network Topology

jason.hurst

Hi everyone,

I have a couple of problems with the results of the NPM Network Topology. It is giving really strange results, like claiming that a server has an unknown interface and it is connected to a core router when I can trace the cable and show that it is not.

It also doesn't seem to recognize a Cisco 4900m as having switch capabilities, so no device that is connected to a 4900m is showing up on NPM Network Topology.

Is there some configuration setting change that I need to make?

Thanks everyone!

Sincerely,

Jason

FormerMember

We had this issue in 10.3.1. While I can't say that our rebuild of NPM on 10.4 was a direct fix for the problem, they do state that their are topology polling fixes in 10.4.

Also, you need to make sure that at least Layer 2 topology polling is turned on in "List Resources".

Even further than that, if you are not in control of the SNMP configuration on your network gear, you may not be aware that the SNMP view has been limited. One of the double-edged swords with Cisco SNMP is that you can fine tune the configuration to only respond to a very select set of OIDs. Sometimes this is done for security reasons, sometimes this is done for performance.

I know that we have some 6500s in our environment that handle a large number of public IPs which fills our ARP tables up into the 30k+ range. An ARP table this large will cause the SUP CPU to spike while NPM is performing it's Layer 3 topology mapping. While this won't impact router performance (as SNMP is a lower CPU priority) we did notice that anything being monitored with SNMP that passed through those 6500s had gaps in polling data. When we turned L3 topology off for those devices, CPU went down to nothing and all other monitoring came back to normal.

Now, we could've implemented an SNMP view to prevent NPM from even seeing the ARP table, but it was just easier to turn L3 off for that device.