BSCI - EIGRP question

Unanswered Question
Jan 18th, 2008

While I was reading through the BSCI Cert Guide from Cisco press I came across the part where they say that certain topologies can pose problems for EIGRP, in particular hub and spoke ones. Why is this the case? Does it have to do with how the query and reply method works? If so isn't this easily fixed by using stub networks, summarization or filtering? Am I missing something here?

Thanks!

I have this problem too.
0 votes
  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 0 (0 ratings)
Loading.
mohammedmahmoud Sat, 01/19/2008 - 01:18

Hi Chris,

I believe that the book was referring to the split-horizon issue with EIGRP, with EIGRP split-horizon is never disabled by default (unlike RIP which disables split-horizon by default on physical FR interface), and thus you must consider this when using hub and spoke topologies.

BR,

Mohammed Mahmoud.

Istvan_Rabai Mon, 01/21/2008 - 13:03

Hi,

To be more specific with this split-horizon issue I tell you a config:

You have router A as the hub router with a serial interface, encapsulation frame-relay, one-one PVC to routers B and C respectively.

By default it will act as a point-to-multipoint interface, split horizon enabled,

You have also routers B and C as stub routers with serial interfaces, encapsulation frame-relay, having one PVC each to router A.

When router A receives an update from router B through its serial inteface, router A will not send these routes to router C because of the split horizon rule: it will not advertise routes to the serial interface through which these routes were received (from router B).

For the same reason router A will not advertise router C's routes to router B.

If you want routers B and C to have each other's routes, the solution to this is to disable split-horizon on the serial interface with the "no ip split-horizon eigrp as-number" command.

This is very tricky, but works.

Cheers:

Istvan

opteronguy Mon, 01/21/2008 - 15:23

That explains it for me. Makes sense.

If you had dual hub routers with each remote site connected to each one, this would then mitigate the problem, correct?

Also, what if you were just injecting a default route back to those stub locations? Wouldn't that again reduce the problem?

Thanks for the replies!

Edison Ortiz Mon, 01/21/2008 - 15:44

1) dual hub ? You mean 2 point-to-point routers?

Yes, it wouldn't have a problem on point-to-point connections.

2) EIGRP stub can also be a workaround.

Another workaround is creating sub-interfaces on frame-relay instead of using the main interface for connection to all spokes.

HTH,

__

Edison.

Actions

This Discussion