[Asterisk-Users] Bonded ethernet ports and *

[Rich Adamson]

> > A fairly common assumption is the failover happens in xxx milliseconds, but
> > due to nic card design (etc) a different MAC address is used in the failover
> > condition. That confuses the hell out of the layer-3 boxes and negates the
> > value of the failover. (All documentation, etc, is correct but actual
> > implementation in this example is limited by the nic card's inability
> > to use a different MAC address from what's programmed into it. There are
> > a large number of current nic cards like that.)
> > 
> 
> I can see this being a problem (not only with layer 3 boxes, but even 
> with other IP endpoints in the same segment, after all, their ARP 
> binding will become useless in this case..
> But would you have any example of this ?

I've never tried to keep a list of nic cards (and/or drivers) that do/don't 
support it. I know some of the Intel cards don't.

> Most of the time, should be a no-brainer to have the NIC use a 
> "non-burned-in-address", in fact, I wonder if is not required by the 
> 802.3 specs (LAA), since protocols like DecNET, SNA and etc tend to 
> require this.

Don't know about the spec requirement; could be now.

Is Decnet and sna still around?  ;)  Sort of like token ring... I still
have to keep an old laptop around that actually has support for a
token ring pcmcia card just in case we run into a client that still
uses it. Haven't had to use it for a long time though. :)