NIC errors and teaming

[Simon Jameson_1]

I have been invesitigating errors that are appearing on the NIC reported by insight manager 7
the server is a proliant 6500 w2k sp3 running 2 NC3123 nics teaming for fail-over.
I ran a batch file to generate network traffic between 2 server and noticed in hours 43000 MAC Recv Errs on the card.
I unteamed the nics and re-ran the batch file and 3 hours later there was no errors at all.
I'm using latest version of teaming (7.11.711.1) and nic drivers (6.1.3.0)
Can anyone shed some light on these types of errors and what might be causing them.
thanks..

[Theo Hill_1]

Which slots are the NIC's in?

[Theo Hill_1]

I guess I could suggest making sure the NIC's are in slots 2 and 3.

[Simon Jameson_1]

they're in slots 3 & 4. What difference should this make??

[Glenn Weavind_1]

Are your NICs teamed for fail-over with the heart-beat function turned on? If so, and they're connected to an Ethernet switch, the heartbeat packets from the (presently) standby nic cause the mac table in the switch to flip, as the switch suddenly 'sees' the same mac address that is in use by the active nic appear on another switch port. Depending on how your Ethernet switch ports are set up this can cause real problems that might cause your rec'v mac errors. Bottom line: in a switch Ethernet environment, when teaming for failover, turn OFF the heartbeat function. Yes you lose some failure detection, but at least it all works correctly.

[Simon Jameson_1]

thanks for the explanation. We use layer 2 switching ethernet network. I've also noticed this problem on servers with no teaming. I've been looking for a correlation between types of cards, O/S, driver versions etc but have found nothing really. Any suggestion as to why this might be??

[Glenn Weavind_1]

Until you mentioned that you've seen this on other servers, I would next have suggested that it might be a faulty switch port - I've seen this. What does the switch supplier say in their support docs about MAC Recv Errs, and what causes their stats counter to increment this total?

[Simon Jameson_1]

the strange thing is that there are no errors reported by our cisco switches on the ports with the servers. Only on the stats reported by insight manager. these are only recieve errors on the NIC and this problem is happening on other random servers with different OS versions and nic's. I can't seem to find any help about what a MAC recive error is from compaq.