Junos OS

Hi,

I have a strange problem, I have two i-BGP peers and I'm trying to export routes from on to the other:

I can see the route that I wan't to advertise in my table:

10.10.10.0/24       *[BGP/170] 2w2d 16:31:26, localpref 100
                      AS path: I
                    > to 10.1.1.1 via reth0.100

My export-policy:

policy-statement test-export {
    term 1 {
        from {
            protocol [ bgp ];
            route-filter 10.10.10.0/24 exact;
        }
        then {
            next-hop self;
            accept;
        }
    }
    then reject;
}

But advertising-protocol don't show that my route i advertised, I have more policys to the same peer but where the routes come from ospf and not bgp that is advertised correct.

Ideas what's causing junos not to advertise bgp-bgp routes?

Hi,

BGP shouldn't advertise a route from one iBGP peer to another. These routes should be advertised by "source"-peer to all others by it's own session. You do have a full mesh, right?

Best Regards

Ulf

hmm...

It's not a full mesh, it like:

router(source)   ---- router2   ----- router3 (this is where i wan't to advertise to)

Hi,

BGP doesn't care that much about physical topology, but for loop avoidance (and maybe other reasons) the BGP RFC 4271 (http://tools.ietf.org/html/rfc4271#page-84) clearly states "When a BGP speaker receives an UPDATE message from an internal peer, the receiving BGP speaker SHALL NOT re-distribute the routing information contained in that UPDATE message to other internal peers" which in turn requires a full mesh, so "router(source)" and "router3" need to have a BGP-peering/session to exchange this routing information.

Ulf

Thank you for a fast reply 🙂

I will change to eBGP between r2 och r3 instead.

This is called the split horizon rule.

You can overcome this by setting internal BGP speakers as route reflectors. If you have lots of BGP speakers in your network it makes more sense to do this, but in your case it probably isn't necessary 🙂