force traffic through a t1 or t3 by ip address, eigrp

[preston]

I have 2 3660 in 2 locations, the locations are connected by a t1 and a t3, running eigrp. I want certain subnets to ride the t1 by defualt, and only ride the t3 in the event of the t1 going down.

for example,

3660A 3660B

192.168.1.x rides ds3 3.x

192.168.2.x rides t1 4.x

If the ds3 would go down, 1.x would go onto the t1, if the t1 goes down, the 2.x would go to the ds3

Again I am running EIGRP, so it wants to send everything through the T3, while I only want certain networks going through it.

If anybody has an example or an idea on the best way to do it, I would appretiate it

[konigl]

Sounds like "policy-based routing" (PBR) is what you want. Search the Cisco website for that phrase or acronym and you should find all the information you need.

I used PBR recently to forward all Internet-bound traffic through a firewall, except for traffic from known internal IP videoconferencing equipment which was allowed to communicate directly with the Internet. Sounds like you would have a similar exercise in specifying "these source IP addresses leave through THIS interface, while those source IP addresses leave through THAT interface".

Make sure you cover all your bases for the returning traffic, too.

[preston]

I am with you there, although aside from 2 networks I mentioned at the core I have about 12 Framerelay sites with a pvc going to each of 3660,s so EIGRP is king.

How does PBR interact with EIGRP? Do you know of papers discussing PBR interaction with EIGRP?

[aretana]

PBR doesn't interact with EIGRP (or any other routing protocol for that matter). PBR overrides any dynamic routing decision. Only if you use the 'default' keyword will the router look into the routing table before switching the packet.

The main thing to remember when using it with a dynamic routing protocol is that PBR is configured (per-interface) per-router. In other words, the consistent routing view that a dynamic routing protocol gives you is altered at specific points in the network. Of importance is to keep an eye on possible routing loops in failure conditions.

The extent of the risk of altering your routing policy only at some points in your network depends very much on the extent and location of the PBR configs. In your case, it sounds like the traffic that you are using PBR to redirect can only possibly be originated on one side of your 3600 routers (the "local" side). Applying PBR only on these interfaces should not cause any real harm.

On the other hand, if you want to extend the "routing exceptions" across the FR cloud, for example and affect multiple hops...that's when special attention to the behavior of the flows in steady state vs possible failure conditions should be taken.

In a future IOS release (potentially late this year), we will be providing a feature called Multi-Topology Routing (MTR), which will allow you to dynamically (using EIGRP, for example) define routing topolgies based on specific policies -- including "traffic from a specific set of subnets". :-)

Hope this helps!

Alvaro.