[c-nsp] Catalyst LAN Input Errors Query...

[Howard Leadmon]

> On Thu, 2008-11-06 at 14:24 -0500, Howard Leadmon wrote:
> > FastEthernet9/48 is up, line protocol is up (connected)
> >   Hardware is C6k 100Mb 802.3, address is 0004.de66.8f73 (bia
> >                                                     0004.de66.8f73)
> >     MTU 1500 bytes, BW 100000 Kbit, DLY 100 usec,
> >      reliability 255/255, txload 3/255, rxload 24/255
> >   Encapsulation ARPA, loopback not set
> >   Keepalive set (10 sec)
> >   Full-duplex, 100Mb/s
> >   input flow-control is off, output flow-control is unsupported
> >   ARP type: ARPA, ARP Timeout 04:00:00
> >   Last input never, output 00:00:43, output hang never
> >   Last clearing of "show interface" counters 00:12:47
> >   Input queue: 0/2000/0/0 (size/max/drops/flushes); Total output
> >                                                            drops: 0
> >   Queueing strategy: fifo
> >   Output queue: 0/40 (size/max)
> >   1 minute input rate 9759000 bits/sec, 1396 packets/sec
> >   1 minute output rate 1505000 bits/sec, 1110 packets/sec
> >      1067610 packets input, 920823086 bytes, 0 no buffer
> >      Received 0 broadcasts (0 multicasts)
> >      0 runts, 0 giants, 0 throttles
> >      980 input errors, 0 CRC, 0 frame, 0 overrun, 0 ignored
> >      0 watchdog, 0 multicast, 0 pause input
> >      0 input packets with dribble condition detected
> >      839374 packets output, 146203703 bytes, 0 underruns
> >      0 output errors, 0 collisions, 0 interface resets
> >      0 babbles, 0 late collision, 0 deferred
> >      0 lost carrier, 0 no carrier, 0 PAUSE output
> >      0 output buffer failures, 0 output buffers swapped out
> >
> > Notice that in less than 15 min I have almost 1000 input errors, but
> > the other more detailed counters show nothing.  I have had the cable
> > swapped, and the LAN card in the PC swapped, still the same results.
> 
> Well, a thousand errors may sound like much, but it's less than 0.1% of
> the total number of packets received.


 Understood, and I think the clients network is working OK, but when all the
other interfaces are running without a constant stream of errors, it has to
make you wonder!

 
> > What is just an input error?   Is this bad hardware, something I
> should
> > just expect on some interfaces to PC's, or what?
> >
> > I have googled around a bit, looked on Cisco's site, and everything
> > says that the input error counter is just the combined count of the
> > other counters like CRC, overrun, and so on, but they are all 0 for
> > me..
> >
> > Any clues on where to look or what would cause this???
> 
> What type of card is it? If you have an oversubscribed path to the
> backplane the switch might drops packets there. AFAIK there's no
> surefire way to find out though.


Basically it's a BSDi based firewall (they need to replace at some point),
that has a pair of Intel Pro/100B adapters installed in it for the in/out
paths.   Both are running 100/FDX, verified with ifconfig, and of course as
you could see from my original posting the switch ports are also 100/FDX.

Just FYI, cables and network cards replaced on the server, but same thing.

 
> Input flow control might help reducing lost packets if they're caused
> by
> oversubscription / too small buffers. This assumes the server NICs know
> flow-control of course.
> 
> Do you have any interface on a similar module with similar traffic/load
> patterns that is not experiencing these errors?

As stated above, it's the two PRO/100 cards generating errors to the switch.
There are other machines/devices plugged in to the various ports that seem
to be working fine, why at first I figured maybe some wonky hardware.

On the issue of traffic loading, and oversubscription.  I don't know what
the max on a WS-X6348-RJ-45 board is, I know it's not the star champ of the
6500 line, but if you look at the data flows the sucker only sees 6-10 meg
of traffic, in fact nothing on that board is pounding the heck out of it, so
I wouldn't think a couple meg of traffic (it was only running 3meg when I
took the samples with the increasing errors) would blow out any port on a
switch like that, but maybe I am wrong..


 
> Regards,
> Peter




---
Howard Leadmon