Re: [Int-area] Fwd: New Version Notification for draft-nordmark-intarea-ippl-01.txt

[Erik Nordmark <[e-mail]>]

On 11/5/15 1:28 AM, Mikael Abrahamsson wrote:
> On Thu, 5 Nov 2015, Dave Thaler wrote:
>
>> I notice that currently RFC 4389 is not referenced at all.
>> It’s Experimental since there are many ways of proxying and it just 
>> describes one of them,
>> but I still think it bears referencing informatively, especially 
>> since section 6 of that RFC is
>> very relevant to the discussion on loops in the document.
>
> I have in live networks running spanning tree and HSRP seen issue with 
> the L2 mac learning timeout being 300 second (cisco default) and arp 
> timeout being 4 hours (also cisco timeout) where in certain topologies 
> the L2 switch would not see one of the routers mac address until the 
> arp timeouted, and flooded all traffic to that router to all switches.
>
> The problem was "fixed" by turning down the arp timeout to 270 seconds.
>
> THis is not exactly what your draft describes, but it might still be 
> worth mentioning?

Mikael,

FWIW I've seen similar interaction with VXLAN where there is dataplane 
learning in the underlay and the overlay, both interacting with ARP timers.

But I don't see how to fit in a discussion about MAC learning in the draft.
Any suggestions on how to fit it in?

Thanks,
    Erik