Route Redistribution

dppagc

Does it make more sense to redistribute OSPF into BGP or the other way around?

Why do people do both? Wouldnt it be very confusing?

late_collision

Short answer, it depends.

If you redistribute a full BGP internet routing table into OSPF, you will have stability problems. However, redistributing OSPF into BGP can be done for many reasons, including public network advertisement or injecting routes into your core.

BGP is also used often to bridge between other IGP's. Lets say your office in LA runs EIGRP, your office in Houston runs OSPF, another office in Atlanta runs EIGRP, and somewhere else runs RIP. You might redistribute your IGP's into a BGP core to connect all of these offices. This gives your network some hierarchy. In order for these offices to know how to reach each other, you might also have to redistribute from BGP back into each IGP, typically using either a default route, a corporate default route, and/or route filtering.

Other reasons may be logical separation of departments within an organization or peering with an extranet.

No method is good/bad, just a tool for a different use case.

So, again, the answer is, it depends.

Hondabuff

I have 2 BGP ISP circuits that come into our company for HA. We run 50 Branch offices with VPN's and use OSPF on the internal network. From the internal network you have to redistribute the OSPF routes into the BGP and vice versa for network connectivity. That's our Sr. networks Engineers baby and no one sure of the voodoo that he used to make it work. Having NP knowledge, I have looked at the configs before and said "nope" not touching it. It has layered configs on it that were never cleaned up and its scary to look at. My side of the NOC we use BGP on +1400 HUB routers and they come back in through the edge routers and redistribute back into the OSPF core. Only reason we didn't use OSPF on the edge and set them up as a stub is because the routers only support RIP and BGP. We chose BGP.

hurricane1091

I don't think I can really add much to what's been said but as an example we are also running one of the interior routing protocols at HQ and at all the branch offices. This gets redistributed into BGP over the MPLS connection. I did lab this up once awhile ago but let me try to see if I still understand correctly and maybe help you.

If you are running EIGRP and have 192.168.1.0/24 at office A in NY, 192.168.2.0/24 at office B in CA and 192.168.3.0/24 in FL and have an MPLS connection, you can re-distribute EIGRP into BGP and advertise that out on the edge router, and you can re-distribute BGP into EIGRP so all of the other routers at the sites have the routes.

Now, you have to make route-maps and set the metrics - which I did once but only in a very simple lab. I couldn't go into crazy details, this whole thing gets a little complex.

You can also have multiple service providers running into the office and re-distribute internet BGP routes into EIGRP/OSPF so the core routers know which provider has the better route.

This is an interesting topic and I'm actually excited to get back to the CCNP after I finish my classes and learn more about it. I wish I knew more to help you out but all I can really post is what I already have. I have some understanding of it but I do not touch this in production, I only made a simple lab once to see how it worked.

EdTheLad

Redistribution in general is bad practice, other methods should be used if feasible. For example, instead of redistributing your igp into bgp, create an aggregate to null and use the network statement to advertise into bgp. If possible on the border router, rather than redistributing bgp into ospf, advertise a default into ospf "default-information originate always", if you have two gateways advertise the defaults with different costs etc..

joetest

EdTheLad wrote: »

Redistribution in general is bad practice, other methods should be used if feasible. For example, instead of redistributing your igp into bgp, create an aggregate to null and use the network statement to advertise into bgp.

Care to expand on this? When is this feasible compared to redistribution between them? (only when you actually can make a proper summary?)

james43026

EdTheLad wrote: »

Redistribution in general is bad practice, other methods should be used if feasible. For example, instead of redistributing your igp into bgp, create an aggregate to null and use the network statement to advertise into bgp. If possible on the border router, rather than redistributing bgp into ospf, advertise a default into ospf "default-information originate always", if you have two gateways advertise the defaults with different costs etc..

I'm very curious how this would work. Care to elaborate on how this would work?