[j-nsp] MX204 and IPv6 BGP announcements

[Mark Tinka]

On 2/6/24 18:48, Lee Starnes via juniper-nsp wrote:

> Hello everyone,
>
> I was having difficulty in getting an announcement of a IPv6 /32 block
> using prefix-lists rather than redistribution of the IP addresses in from
> other protocols. We only have a couple /64 blocks in use at the moment but
> want to be able to announce the entire /32. In cisco, that would just be a
> holddown route and then announce. Not sure how it works to Juniper.
>
> I configured a prefix-list that contained the /32 block in it. Then created
> a policy statement with term 1 from prefix-list <list> and then term 2 then
> accept. Set the export in BGP protocol peer of this policy statement and it
> just ignores it.
>
> Now this same setup in IPv4 works fine.
>
> After a week of going round and round with Juniper TAC, they had me setup a
> rib inet6 aggregate entry for the /32 and then use that in the policy
> statement.

This is the equivalent of the "hold-down" route you refer to in 
Cisco-land. Useful if the route does not exist in the RIB from any other 
source.

I'm guessing your IPv4 route just works without a hold-down route 
because it is being learned from somewhere else (perhaps your IGP, iBGP 
or a static route), and as such, already exists in the router's RIB for 
your export policy to pick it up with no additional fiddling.

Typically, BGP will not originate a route to its neighbors unless it 
already exists in the routing table through some source. If it is an 
aggregate route, a hold-down pointing to "discard" (Null0 in Cisco) is 
enough. If it is a longer route assigned to a customer, that route 
pointing to the customer will do.

Mark.