[j-nsp] Where and how to inject the default route in DFZ

Stepan Kucherenko twh at megagroup.ru
Fri Nov 27 03:09:50 EST 2015


I just use a generated 0/0 route which is active only if I receive 
specific prefixes from upstream(s).

If you don't want 0/0 in FIB then just add no-install. Not perfect but 
better than no delay at all.

I wish I could say something like "thos route is active when there are X 
routes received from this neighbor(s) and they're already in FIB" but I 
didn't find anything like that either.

On 26.11.2015 18:30, Mark Smith wrote:
> Hi list
>
> This is best explained by an example.
>
> Router R1 has a full bgp table (~550k prefixes). R1 needs to announce a
> default route using OSPF and BGP. The worst issue is when R1 boots up.
> Assume there is a static 0/0 route to discard. R1 brings up OSPF
> adjacencies and starts announcing 0/0. Blackhole. Next R1 brings up BGP
> adjacencies and starts announcing 0/0. Another blackhole. Only after R1 has
> received the complete route table (and pushed all prefixes to FIB) will the
> blackhole disappear.
>
> What's the best solution to work around this? Where and how do you generate
> the default?
>
> With OSPF or ISIS one can use overload w/ timeout. It works.
> What about BGP? out-delay? One can use "routing-options generate route ..."
> for default. But what prefix to use as contributing route? One can announce
> a dummy prefix from e.g. RR, but still this does not guarantee that R1 has
> all necessary routes (in FIB) before announcing default. And this probably
> leads to transient routing loop while R1 has default pointing to e.g.
> loopback of RR, but routers in between see the OSPF default announced by R1.
>
> Overload timeout + generate default?
>
>
> Thanks
> _______________________________________________
> juniper-nsp mailing list juniper-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/juniper-nsp
>


More information about the juniper-nsp mailing list