Lots of input errors...

[Mike Hoskins]

On Wed, 25 Jun 2003, Shawn Ramsey wrote:
> > netstat -I xl0 -w 1
>            input          (xl0)           output
>   packets  errs      bytes    packets  errs      bytes colls
>      6918    22    8525822       5631     0    2770466     0
>      7317    21    9262852       6041     0    2696855     0
>      7839    26   10090955       6426     0    2686936     0
>      7260    14    9318261       5789     0    2407180     0
>      6653    14    8255322       5552     0    2693452     0
>      7818    17    9966908       6323     0    2693943     0
>      7003    12    9056270       5406     0    2250436     0
>      7104    17    8904400       5963     0    2815142     0
>      7287    12    9185995       5937     0    2747249     0
>
> This seems excessive. What are the likely causes of this other than say
> a bad cable or switch ?

Improperly negotiating 100-BT/FD and generating lots of late collisions,
for one.  Is the switch managed?  What does it's syslog output or the
local CLI say about the port(s) in question?  In Cisco parlance, you may
want to clear the interface counters and observe 'sh int...' output while
transferring a large file.  I must say, however, that if negotiation is to
blame (and Cisco's is notoriously bad), you should be seeing degrading
network performance.  (I think you'd notice that.)

> I believe the same thing was happening on our
> other interface when we had this much traffic going into it, and its
> plug into a different switch entirely.

Is it also xl0, and connected to the same brand of switch?  One thing to
try if you rule out other issues (if the server isn't too busy to allow
it) -- throw in another (non xl) NIC.  I haven't used xl* in awhile.  I
doubt it's a driver issue, but swapping NICs may rule it out with
certainty.

-mrh

--
From: "Spam Catcher" <spam-catcher at adept.org>
To: spam-catcher at adept.org
Do NOT send email to the address listed above or
you will be added to a blacklist!