[c-nsp] Snmp monitoring of 10GigE Interfaces

[Justin M. Streiner]

On Thu, 10 May 2007, Kumar Dasari wrote:

> I am having rather erratic and inconsistent results for bandwidth usage 
> reports from different SNMP software packages (MRTG, Cacti, Solarwinds 
> etc) when monitoring 10GigE interfaces on Cisco 7609 Routers. For 
> example there is discripency in what the snmp software says what the 5 
> min bps output rate is, and what the show int te1/1 shows on the router 
> itself. Software is lower always. Any suggestions on how I can fix this?

On average, how far off are the SNMP counters from the output of a 'show 
interface TenGigXXX'?

A few thoughts off the bat...
1. Are you running a version of software on these routers that has an SNMP 
bug?
2. Are you polling the 64-bit counters for your 10 gig interfaces?
3. What is the load-interval set to on the interfaces?  If you don't see a 
"load-interval XX" under specific interface configs, then it's set to the 
default value, which I believe is 5 minutes.
4. Are the graphs always lower, even in the 5-minute/daily traffic views?
MRTG will wash some of the traffic peaks out of the graphs over time, 
unless you specifically tell it to preserve them, but you wouldn't see 
this until you get into the longer-term views.

jms