[c-nsp] OFF TOPIC: EIGRP Multicast Flow timer

[Paolo Lucente]

Hi Omar,

I'm far from being an EIGRP fellow - but should be able to confirm your
thoughts: retransmissions of multicast EIGRP updates to slow/unresponsive
neighbors happen as unicasts. Such "exception" is handled by the RTP layer
in order to not affect negatively convercence. The reason being a router
can transmit the next reliable multicast packet only if acknowledgment of
the previous one has been received from all of its peers. If this doesn't
happen, slow/unresponsive peers are handled separatedly (unicasts), whereas
healty peers continue receiving multicast packets as usual. 
Then, if the neighbor is not just slow but totally unresponsive, it gets
removed from the adjacency table after having reached a maximum number of
retransmissions (which usually is a shorter timeframe compared to the hold
time). Hope this helps.

Cheers,
Paolo

On Sat, Sep 08, 2007 at 05:12:11PM -0500, omar parihuana wrote:
> Hi guys,
> 
> Recently, I was reviewing about EIGRP, and found one question that
> unfortunately Cisco web page don't help me, after check in google I found
> the follow explanation (extracted of cisco-nsp archives):
> 
> [ ... ]
> 
> However I have some questions, in accordance to explanation above: if router
> one send to router three a TLV telling not listen to any further multicast
> packet, how router three learn new updates? special unicast?  of course
> router three is UP and working fine again. Could you give more details about
> that?
> 
> Thanks in advanced...
> 
> Rgsd
> 
> -- 
> Omar E.P.T
> -----------------
> Certified Networking Professionals make better Connections!