Comware

Hi All,

We use HP NIC Teaming(TLB with NFT) mode on our Windows Servers. Spanning tree is not in use and no special switch configuration such as Etherchannel, etc is in place as our understanding is that with TLB that is not required.

We have been advised by our Network Support group that they are seeing a large number of error messages on the Cisco Switches. Error message is:

Nov 1 12:04:49 gp37s-v0126.a.b.c.d 177: Nov 1 12:04:48 AEST: %SW_MATM-4-MACFLAP_NOTIF: Host 0030.54a2.1dc7 in vlan 1120 is flapping between port Gi1/0/4 and port Gi1/0/1

This is happening across a good number of hosts and switches.

Could someone please shed some light.

Many thanks,
Saket.

Strictly speaking, I believe this forum is for _HP_ ProCurve switches :)

I suspect you need to configure a "proper" LACP (IEEE standard) trunk or an EtherChannel one (Cisco proprietary) depending on which of the two the teaming stuff supports. My first vote would be for the LACP (Link Aggregation Control Protocol) patch since that is the de jure standard.

That should quiet the switch. Switches really don't like seeing MAC's move around all that much. Modulo actual trunking/aggregation/bonding/teaming enabled on the switch, it will (IIRC) only send traffic destined for a given MAC out one port. So, while you may be getting outbound balancing on your server(s) I suspect you may not be getting all that effective inbound balancing.

This is strange.. the MAC should not be moving around a lot.

With standard settings (TLB) NIC teaming sends packet from all team ports, but with the original MAC-addresses. So you should see separate MACs on each port, but with the same source IP.
Did you change the default TLB settings? Which particular settings are you using?

NFT will move the mac, but only when there is a failover, which hopefully doesn't happen that often.

You should, as you say, not need any special config for TLB.

Rick, LACP or Etherchannel configuration on Switches is only required when you setup the SLB mode on the host end.

As per the original question, the NIC teaming mode in use is TLB - this requires no special switch configuration - yes spanning tree is turned off...!

So, I am unsure why the MAC seems to be moving on the switches and the error messages...

Thanks,
Saket.

WRT TLB, I'll confess to being more Linux-literate and going on incomplete intuition. Still, it sounds like TLB isn't going precisely what TLB is supposed to be doing. It could be a long-shot, but is there any chance that the individual ports on the NIC(s) got mistakenly configured with the same MAC?

What/where is the chapter and verse on TLB for Windows?