Re: [nsp] Wrong interface counters on 2948G-L3?

From: sthaug@nethelp.no
Date: Thu Jan 04 2001 - 18:21:13 EST


> Dunno, but one side has a large amount of multicast activity and the
> other doesn't, one side has a large number of unspecified output
> errors. I'd say unless you can account for that, you can't expect
> the other numbers to really add up...

I have every reason to believe the apparent large amount of multicast
activity is incorrect. For one thing, it claims more multicast packets
than total packets input:

> 75179 packets input, 46122114 bytes, 0 no buffer
> 0 watchdog, 97436 multicast

I have done some experimentation on the unspecified output errors. It
is clear that these errors are related to packet size, and presumably
then also to the VLAN encapsulation used (ISL in this case). Using
1482 byte packets (instead of 1500 byte) results in no errors. Using
1484 to 1500 byte results in one error per packet - but the packets
clearly get through to the other side and are valid IP packets (e.g.
1500 byte pings are received by the other side and returned with no
packet loss).

Thus I believe it's clear that this "problem" is a result of the
2948G-L3 miscounting certain types of packets.

Steinar Haug, Nethelp consulting, sthaug@nethelp.no



This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:12:24 EDT