Cisco Support Community
Showing results for 
Search instead for 
Did you mean: 
New Member

incorrect bandwidth values via snmp?

I have two cat 2900 16 port switches, for some reason when i read the bandwidth utiliz. values via snmp (PRTG, and MRTG give same results) i only show a steady 1-1.5 kbps up and down flow with occasianal spikes of the upstream to 2 or 3kbps..this is the same across all ports, now if i send something across the switch, nothing changes, nor is any other traffic i throw at the switch reflected... the peculiar thing is if i access the switch's web interface while conducting a 10.5 mbit transfer across the switch, i will see a spike of correct values (as i read on the hosts machines) reported by the snmp agents but only for 2 or 3 reads across 5 second periods..then its back to what i described again.. I've done factory resets on both and both of these snmp agents correctly read the bandwidht values from my dell and intel switches so i've ruled the agents out..

any ideas?

thanks, joe


Re: incorrect bandwidth values via snmp?

We once had an issue with importing SNMP data where the PC used a financial format to display the numbers. This format gives a dot per three digits which can be misinterpreted as a decimal point by the agent. I do not know exactly how you are importing your data but this perhaps explains why you are not getting beyounf the kbps range.



New Member

Re: incorrect bandwidth values via snmp?

Thanks for the quick reply but thats not the problem. its not a units issue as i see somekind of traffic but when i'm maxing out the 100mbps port i see no change in the snmp reported graphs..but like i said i will see the actual values if i'm reloading or access the switches web interface.

this is such a weird issue..and its affecting two swithces that are the same model but from different places.

any ideas?

New Member

Re: incorrect bandwidth values via snmp?

Can you check which OIDs you're polling? I'd expect

ifInOctets.1 - .16 ( - .16)


ifOutOctets.1 - .16 - .16)

The received value should be a (rising) counter, the utilisation has to be calculated by the application.

Maybe you've the opportunity to trace the snmp pdus (e.g. with ethereal)?

CreatePlease to create content