EIGRP Feasible Successors

sultan_j · ‎06-26-2007

We run quite a large multi-site Cisco network using EIGRP as the routing protocol. Many of the sites have a backup link but looking at the topology table there are virtually none that have a feasible successor. This seems to defeat the object a little of running EIGRP, because of instead of immediately switching to the backup route it has to recalculate. It ends up using the backup route eventually anyway!

I know that EIGRP must be doing this because of its AD versus FD calculation to avoid routing loops, but as it eventually uses the backup route it means that it could have used it immediately.

Is it advisable or best practices to alter the delay or bandwidth of links where there is a backup link so that a feasible successor gets put in the topology table? I'm pretty sure that whatever alterations are put into these fields can't result in a routing loop due to EIGRP's calculations (well I'm not 100% sure but I tried it in a home lab and it seemed to be fine).

Can somebody advise me on the best practices please? After all there's no point in running a supposedly very fast converging routing protocol if it never actually inserts any feasible successors into its topology table and so ends up not being particularly fast.

Richard Burts · ‎06-26-2007

JUNAID

I would suggest that the best practice is to leave the bandwidth and delay at their default value. As a Cisco instructor you are aware that the fundamental concept of the DUAL algorithm is that EIGRP should not immediately use an alternate route (feasible successor) unless it is CERTAIN that the alternate path is loop free. If you start adjusting the parameters to artificially create feasible successors where they do not really qualify then you are defeating the calculation of DUAL. At that point you are manually assuming the responsibility of assuring that the paths are loop free. I do not believe that it is best practice to substitute manual processes for automatic processes.

As you say, EIGRP will use the alternate path if it is loop free. It only takes a bit longer to get to that. I believe that the small amount of delay is worth the safety of having an accurate calculation of DUAL.

HTH

Rick

HTH

Rick

royalblues · ‎06-26-2007

I agree with Rick but if you are pretty sure that all the backup routes/routers are downstream, it can be changed for quick convergence.

Narayan