I have edge routers and access routers peering with a route reflector. My access routers run OSPF to all my edge routers. The route reflector holds my public networks but the access routers hold the actual blocks. For the example of the issue i am having i will use block 200.200.200.0/24 as an example. 200.200.200.0/24 block as a whole is in the routing table being assigned to a customer.

The edge router is learning 200.200.200.0/24 from both the access router (OSPF) and the Route Reflector (BGP). The edge routers are preferring the OSPF routes (normal behavior) instead of the BGP routes. This is causing the 200.200.200.0/24 route to not be re-announced by the edge router to the carrier (because it is not active).

What is the best way to get that route to be shared with the carrier. Should I do "Redistribute OSPF into BGP" and then filter the routes? Is there a better way?

Hope that was easy to understand. Thanks!

Why not use one public AS for your edge routers (presumably doing BGP with the outside world), and a private AS for the access routers, and eBGP between them? Then you don't need route reflection. Or do the access routers need to peer with customers over the public AS?

Correct.. the access routers are also running bgo for customers to peer with.

I also have a total of 8 upstream peers and 3 downstream. That's why I am building the route reflectors. To improve scaling.

Why not use one public AS for your edge routers (presumably doing BGP with the outside world), and a private AS for the access routers, and eBGP between them? Then you don't need route reflection. Or do the access routers need to peer with customers over the public AS?

This is the same issue you described in viewtopic.php?f=14&t=98095&p=698166#p698166

Correct.. the access routers are also running bgo for customers to peer with.

I also have a total of 8 upstream peers and 3 downstream. That's why I am building the route reflectors. To improve scaling.

OK.. well then as a thought, why not use a routing filter to set the distance for the route? That way you could make the BGP-learned route active and it should reflect.

I need the Edge router to prefer the OSPF because I do not want to route traffic through the reflector.

I need the Edge router to prefer the OSPF because I do not want to route traffic through the reflector.

Could you simply establish additional peerings between the access and edge routers besides the reflector? Having a reflector doesn't necessarily mean that everything has to be peered only with the reflector. How many routers are you talking about here?

I think in general it would be preferable to set up extra peerings if necessary rather than relying on redistribution.

We have over 10 edge routers. We are switching to route reflectors because a full mesh is too taxing on the 1072s. Turning up more peers also wont help share the BGP routers when the OSPF routes override the BGP.

For now I have "Redistribute OSPF" into bgp then filtered the crap out of the outbound BGP so the OSPF routes do not get shared with the route reflector.

HI! I think that the best solution for you will be to work with both protocols in a recursive way, I mean, you'll need to set up OSPF for convergence and loopback distribution only, it won't announce any public IP on it, so you can set up an iBGP session between each router to the RouteReflector (using loopbacks only), and you can advertise your public IPs using BGP only; it will announce your public subnets over your entire network, and it won't force traffic trough the RouteReflector (except that the OSPF is miss configured).

The IPA guys are expert deploying this solution.