Please see the attached Image of the proposed new network map. All IP's are ficticious....Right now we are running basiclly the same map, but with only 1 border router. We are adding a second border to take load off the first and, most importantly, protect against failure of one of the border routers.

Im trying to achieve total redundancy, against a border router failure. This happened this past week, which is why we are doing this.

Take a look at my map......

Let me give you a scenario:

Lets say we have customer coming out of core3 on a network 200.200.12.0. Lets say border2 has some sort of an issue, where all internet gets cut off but ethernet to backbone is still up. The BGP advertisments from border2 to the internet providers will drop. Ok...thats good. But how will core3 know to drop that route to border2? Ebgp on border2 is working and not advertising any of our networks anymore since the links are down, but core3 will still try to hit border2 not knowing it cant get anywhere.

I had a few ideas. Please see attached network map file....

A. Tell all my ebgp peers to just send me a default route. Then run bgp on the cores and establish a bgp connection between the cores and the borders. This doesnt seem like the best way to do this.

B. Run OSPF internally and some magic will work? Ive never run OSPF before, so ive been reading about it. I guess OSPF in the cores would have to send the customer network routes out to the borders...which would then be injected into BGP (Vs the static advertisments I do now)?

Ideas? Thanks in advance guys. I have all the equipment coming this week to upgrade the network.

--John

Really that is more of a Cisco question

Run iBGP between the border routers. Run OSPF to the external network. Use the "default-information originate" command without the "always" keyword. That will cause the Cisco router to advertise a default route to your OSPF network only when it already has a default route in the routing table (which would be learned via BGP). BGP drops, that default route via BGP drops, it withdraws the OSPF default route, all traffic goes to the other border router.

fewi, thanks for the advice. Im going to set up a test next week to give it all a shot.

--John

Oh I forgot, keep the static BGP advertisements. While it is possible to redistribute OSPF into BGP in a fairly static network like yours I don't think that is a good solution. You would have to take precautions to make sure BGP advertisements aren't withdrawn every time a customer router reboots, for example. Static is simple. Simple is good.

Just run OSPF across the backbone between the border routers and the core routers, and use the border routers to insert default routes into the OSPF area. Run iBGP between the border routers. Core routers just send Internet traffic to the closest border router based on the default routes they learned via OSPF. That border router then either sends straight to the Internet, or to the other border router if that has a better route and advertised it via iBGP - if you do full BGP tables, that is. I'm assuming you are, with five upstream peers it'd be worth it.

Wouldnt have to worry about customer reboots causing OSPF--->BGP changes, since i hand out the network on the mikrotiks, with the tik as their default gw. So unless my core reboots nothing would change...

I see whtat you are saying. Since the borders are running ibgp...if one suddenly loses its internet links, it will send traffic via the other. If an entire border router dies, then OSPF will just send to the other border. Correct?

That makes sence to me...

--John

Exactly.

Thanks man, i appreciate the help. Ill update the thread next week when i start setting this up.

--John