Brief Wrap-up on the Lab from the Last Few Weeks

EIGRP – How to Force a Query with Two Routers

This brief post is a wrap up of the lab I posted a few days back. The question is this: is a 2-router, 2-serial link design enough to experiment with EIGRP metrics and with the concept of when EIGRP needs to send a Query, and when does it not need to send a Query?

If you didn't read the last few posts, this is really #3 in a series of 3 related posts, so you may want to start here with the RIP lab, and then here with the EIGRP lab.

The solution to this problem, beyond the IP addresses and bringing up the interfaces, is relatively straightforward. R1 and R2 both need EIGRP enabled on all 3 interfaces, and R2 needs no auto-summary configured for the same reasons it was needed in the RIP lab (to make R2 advertise 172.16.1.0/25 instead of 172.16.0.0/16). So, picking up the configuration in the RIP lab solution, removing RIP, and adding this config, gets you most of the way there:

R1:

Router eigrp 1
 Network 10.0.0.0

R2:

Router eigrp 1
 Network 10.0.0.0
 Network 172.16.0.0
 No auto-summary

At this point, R1 should have 2 IP routes for 172.16.1.0/25 - one out S0/0/0 (top link), and one out S0/0/1 (bottom link). The lab asks that with all links up, that R1 has only a single route, through the top link. This requires straightforward metric manipulation. With EIGRP, that means changing bandwidth or delay on one of the interfaces that matter. So, which ones matter?

  • Top route: R1's S0/0/0 and R2's Fa0/0
  • Bottom route: R1's S0/0/1 and R2's Fa0/0

Note that both routes' metric calculations are impacted by R2's Fa0/0 interface, so changes to delay or bandwidth there change both routes, and change them by the same amount. So, either you need to make R1's S0/0/0 better (higher bandwidth or lower delay), or R1's S0/0/1 worse (lower bandwidth or higher delay). For example, instead of the default delay of 20,000 microseconds, you could change R1's S0/0/0 to a delay of 10 microseconds with the delay 1 interface subcommand. That would lower R1's metric out S0/0/0, and make R1 use a single route out S0/0/0. (I know, it's weird - the unit of the delay command's parameter is tens-of-microseconds.)

You can confirm that R1 does not send a Query to fail over to the route through R1's S0/0/1 in a couple of ways. First, you can use a debug command (debug eigrp packet query), shutdown the top interface, and look for messages stating that R1 sent any  Query messages. You can also rely on the facts that EIGRP does not send Query messages if a Feasible Successor (FS) route exists, and that the show ip eigrp topology command only lists successor and feasible successor routes.

With the 2-route topology, the route over the bottom link will always be a feasible successor route if all you did was change the bandwidth and delay to favor the top route. To make the route over the lower link not be a feasible successor, you have to think outside the box a bit, squarely into CCNP territory. You could use an offset list to change the metric reported by R2 to R1 over the lower link, such that R2's Reported Distance (RD) is higher than R1's metric (Feasible Distance, FD) over the top link. You can even think further outside the box, and change the Administrative Distance (AD) of the route learned over the bottom link. If interested, go back to the previous post, and look for the samples from Dan C. of both such cases.

Post questions if you have them. I'll move on to something else for the next post.

Wendell

Join the Network World communities on Facebook and LinkedIn to comment on topics that are top of mind.
Related:

Copyright © 2010 IDG Communications, Inc.

SD-WAN buyers guide: Key questions to ask vendors (and yourself)