BGP Multihoming for two different campuses and with two ISP's

[mazenfawwaz]

I have two core routers for two campuses linked with a dark fiber. Each router is peering with a different ISP and each campus is advertising a seperate network but both networks use the same ASN. I was told that I will need to implement a GRE tunnel between tht two campuses and through the ISP's links in order for the two campuses to be able to talk to each if the direct fiber between them failed.

Isn't it possible that the two campuses use EBGP to talk to each other?

Note: Each campus border router receives a default route from its ISP, so there is no need for one campus to receive the route of the other campus.

[Edison Ortiz]

BGP loop prevention mechanism will drop received routes that contain its own ASN in the AS_PATH.

You can change this behavior with the allowas-in option.

http://www.cisco.com/en/US/docs/ios/mpls/command/reference/mp_m4.html#wp1013322

HTH,

__

Edison.

[mazenfawwaz]

Hi Edison

Thanks for your reply.

But this will not happen in my case as I will only receive default routes from my ISP's.

Is this the only issue?

Take Care,

Mazen

[Edison Ortiz]

Correct, I must have added the Note after I posted since I didn't see it on my original reply.

The issue would be if you are receiving full routes which would include any routes originate from each of your Edge routers.

[nick.mueller]

I believe your assessment is sort-of correct in the situation you specify -- that you only receive default routes from your ISPs.

The only other troubles I could see are:

-If you are advertising your entire IP space from both routers, the router at Site A may receive packets for addresses at Site B, and they will loop between your router and the ISP until TTL expires. The solution is to advertise only the subnets that exist at each site, along with a summary route(s) for your entire space (the summaries will allow you to keep routing correctly if, for example, Site B ISP fails, but your dark fiber is still up).

-You will no longer be able to route to private IP address space inside your organization from site A to site B -- presumably your upstream ISPs will filter those packets, and wouldn't know where to route them regardless.

There could be additional issues, but those are the two that leap out to me.

[mazenfawwaz]

Hi Nick

That was helpful.

Regarding the loop, I am assuming that IBGP will take care of that. Each site will advertise its block to its EBGP peer as well as its IBGP peer. When the dark fiber is up, the border router will advertise to the EBGP peer the local block and the one learned from IBGP. When the dark fiber goes down, each site will stop receiving the IBGP routes and hence will only advertise its own block, which would prevent the loop.

Am I missing something here?

Mazen

[nick.mueller]

If that is how you are doing your advertisements, then yes, that should work.