[nsp] BGP metric?

[Oliver Boehmer (oboehmer)]

> Dear All,
> We have IPv6 setup with 6PE and also with IBGP. We are using 6PE for
> backup now, since we could upgrade only one path only to support IPv6.
> We experienced a problem, that 6PE routes are preferred to native
> routes. 
> 
> cntrl.6net.hbone.hu#show bgp ipv6 unicast 2001:738:6001::/48
> BGP routing table entry for 2001:738:6001::/48, version 525
> Paths: (2 available, best #1, table Global-IPv6-Table)
>   Advertised to update-groups:
>      1          5          8
>   Local, (Received from a RR-client), (received & used)
>     ::FFFF:195.111.97.231 (metric 116) from 195.111.97.231
> (195.111.97.231)
>       Origin IGP, metric 0, localpref 100, valid, internal, best
>       Originator: 195.111.97.68, Cluster list: 195.111.97.231
>   Local
>     2001:738:6000:1::1 (metric 340) from 2001:738:0:3::1
>       (195.111.97.31) Origin IGP, metric 0, localpref 100, valid,
>       internal Originator: 195.111.97.68, Cluster list:
> 195.111.97.31, 195.111.97.7 
> cntrl.6net.hbone.hu#
> 
> 
> As far as I can see the metric difference is in the metric in
> parentheses (116 and 340). Where these metrics come from? This is the
> only attribute that makes the decision? How can I change these
> attributes to make 6PE backup only.
> 

metric 116 and 340 are the IGP metric to the next-hop. With all other
BGP attributes being equal, IGP metric is taken into account. Rather
than changing IGP metric on links (which might break other things) I'd
apply some BGP policy like assigning a higher localpref or weight or
mess with BGP MED/metric in order to administrativly select the route
you favor.

	oli