[c-nsp] igp / ebgp problem ipv6

[Bernhard Schmidt]

Paul Stewart <paul at paulstewart.org> wrote:

Hello Paul,

> We have our first IPv6 block advertising to the world (for quite a while
> now) and have started to actually route some small blocks of it internally
> via OSPF.  Our /32 is advertised via eBGP no problem and the world can see
> it..
>
> Internally, we have a series of /128 loopbacks, /126 point to points, and a
> /64 block setup for some servers.  Obviously all small chunks of the /32
> assignment.
>
> My problem is that the world can reach our border routers but traffic will
> not route beyond the border.  Internally, we can route traffic no problem..

My first guess would be that your inbound traffic is routed just fine,
but your internal routers have no route back. This looks like
blackholing in a traceroute from an external host.

Check both "sh ipv6 route <internalhost>" and "sh ipv6 route
<externalhost>" on all routers in the path, I guess the first one will
return a valid route on all your routers but the second one only on your
edge. You need to tell your internal routers how to get out of your
network. If you only have one edge router you can put a default route
into OSPF. You have "default-information originate" in there already,
this will only work if your edge router really has an exact ::/0 route
in his table. Check for that and add "default-information originate
always" if it doesn't.

Bernhard