Then you can set BGP local preference for a route if you need to select a specific egress point in your AS.

Redistribution is a bad habit to get into when you're just learning routing protocols.
It definitely has its place, but the protocols handle redistributed routes differently than natively originated routes.
(true for OSPF, BGP, EIGRP, and I'm sure IS-IS and RIP)

I recommend that you originiate the routes by adding networks to Routing > BGP > Network
(e.g. /routing bgp network add network=192.168.1.0/24)

Then you could put a filter on R2 where it will do an AS-PREPEND for Network1 when sending to the peer address for Link2, and an AS-PREPEND for Network2 when sending to the peer address of Link1.

This is better because AS-PREPEND is the typical way to engineer inbound traffic. (you could also use MEDs since this is the case of multiple connections between two ASNs.) Furthermore, if Link1 fails, the traffic for network1 can use Link2 as a backup path, and vice-versa.

Any manpage for routing filters ? In the wiki , there are (usefull) examples, but I could not find a reference documentation.
Anyway we have failover so we use iBGP to redistribute BGP learned routes. But for eBGP why not using routing bgp network, I will think about it.

/routing filter
add action=accept chain=link1-out prefix=192.168.2.0/24 set-bgp-prepend=2
add action=accept chain=link1-out
add action=accept chain=link2-out prefix=192.168.1.0/24 set-bgp-prepend=2
add action=accept chain=link2-out

Any manpage for routing filters ? In the wiki , there are (usefull) examples, but I could not find a reference documentation.
Anyway we have failover so we use iBGP to redistribute BGP learned routes. But for eBGP why not using routing bgp network, I will think about it.

Make a filter chain for each eBGP peer (e.g. link1-out and link2-out)
Set the out-filter for each peer as appropriate. Here is a basic way to do the chains:

Code: Select all

/routing filter
add action=accept chain=link1-out prefix=192.168.2.0/24 set-bgp-prepend=2
add action=accept chain=link1-out
add action=accept chain=link2-out prefix=192.168.1.0/24 set-bgp-prepend=2
add action=accept chain=link2-out

These filters allow all prefixes to be advertised, but will prepend the AS-PATH when 192.168.2.0/24 is being sent to link1, and when 192.168.1.0/24 is being sent to link2. (of course, your actual addresses go here). If you need to block any advertisements, you would put rules before the default allow-all rules.

If these BGP sessions have anything to do with the public global BGP table, then you're almost certainly going to want to make "block" become the default action, and only allow your actual network prefixes.