Re: [opennhrp-devel] Problem multicast over tunnel

[Jean-Francois D. <jf...@gm...>]

The problem here is that the outer ip (GRE destination) is the one
from the inner packet dst address (the ospf mcast address), this is
not correct., it should be 172.17.1.X. This can come from 2 things: or
the kernel does not ARP for it (hence nhrpd cannot answer with the
proper address) or nhrpd replies with the 224 address.

J.

2010/1/15 Timo Teräs <tim...@ik...>:
> Jean-Francois Dive wrote:
>>
>> Running ospf over the nbma network provided by gre/nhrp should be
>> transparent to the intermediate routers as we build an overlay network
>> which is completely separated from the carrying one. IOS wise, mcast
>> packets (depending on the config / hub-spoke) will:
>>  -> be sent to a static address (ip nhrp map multicast nbma_address)
>>  -> be replicated to all proto address we have in the cache (ip nhrp
>> map multicast dynamic).
>
> Correct. And that was my point too. (Technically though, the
> replication part is not exactly working yet in opennhrp. I have a
> working patch for it, but it's not in git yet. Currently released
> opennhrp cannot really act as a replicator.)
>
> However, I'm not entirely sure that link-local multicast packets
> should be replicated anyway. I'm currently using BGP only within
> dmvpn, so i'm not sure how the ospf multicast should be handled
> exactly. And my quick memory says that cisco docs give some
> alternatives how to set it up.
>
> But the original mail said:
> "And, as I have another router between the opennhrp computer and the
> remote router, the packets are unable to arrive to the destination
> router."
>
> There's extra router. Multicast won't span extra computers without
> multicast routing. It's not a opennhrp problem, but fault of the
> extra router. Multicast packet's inner destination should never be
> rewritten. It just gets sent to multiple specific nbma destinations.
>
> - Timo
>