[ILUG] billions and billions

[Vincent Cunniffe]

... of errors on a network card?

I've been wondering about this for some time : my Linux server, Heimdal,
generates literally billions of errors on the eth0 interface, without
seeming to suffer from any problems. It's solid as a rock (never crashed
even once, with months of uptime between hardware upgrades and power
cord removals) but the errors keep accumulating. It was literally up to
several hundred *billion* errors when I switched it off yesterday.

The card I removed is a venerable but reliable 3Com 503, which has always
worked perfectly, and the one I replaced it with is a brand new Intel
EtherExpress Pro 100+. As soon as I had the new card installed (which 
worked perfectly : just add module, swap cards, power back on :-), the
error count started climbing again. It had hit several million after a
few minutes.

The odd thing is that this machine has never had a single problem. Last
night it was comfortably shifting 5MB/s, on an uncongested network, with
not even a flicker of collision on the switch, but the reported error 
count keeps rising. This was originally a RedHat 5.2 box, since upgraded 
to a few bits of RedHat 6.0.

So, what causes the errors? Are they actual network errors, 'failure to
deliver' errors, or what?

Regards,

Vin