EIGRP question...

Darden, Patrick S. darden at armc.org
Mon Dec 1 14:24:30 CST 2008


My first thought for this was: route filtering.  My second thought 
was: use different AS#s.  Then I reread your question and thought 
of something far simpler.

It seems to me if you are migrating from provider A to provider B
then you should set everything up for B, then shut down the 
interface to A.  If everything works, then you are good, if not
then you bring that interface back up in a hurry!

Is this more complicated?  Are you, for example, moving to B
but planning on keeping A as a backup?

Or, perhaps your MPLS migration is from non-MPLS to an MPLS based
system?  So you are keeping A and B?

Or perhaps A and B are customers, not really providers?

Anyways, not sure of your exact situation, but to answer your
question directly:

EIGRP uses 5 metrics for weighting paths
--delay
--bandwidth
--reliability
--load
--MTU

It does not use hop count for path weighting.  This is a problem, as it would be your easy solution--pretty much auto-solving the condition.

1.  You can't really count on delay, and you can't fiddle with it in any easy manner.
2.  You could artificially limit bandwidth using your core.  I am unsure if this would help you; however, it would answer your stated question.
3.  Reliability--calculated dynamically.  Not helpful for you.
4.  Load--this is the load on the interface I think (0-255).  Calculated dynamically.  Not helpful for you.
5.  MTU--I have no idea how you could use this.  Not user configurable in this case, I don't think.  I have never used it myself for metrics.  It is theoretically used, but never seen it used (in eigrp).

Not sure if this helps.
--p


-----Original Message-----
From: Mike Lyon [mailto:mike.lyon at gmail.com]
Sent: Monday, December 01, 2008 2:49 PM
To: Nanog Mailing list
Subject: EIGRP question...


Howdy,

So I am working on an MPLS migration from provider "A" to provider "B"
of which both terminate into my core via customer prem routers. I have
a single EIGRP process between my core and the two customer prem
routers supplied to me by both providers, of which I don't have access
to. My question is, I would like to take the routes that come in from
the neighbor "A" router and apply some kind of metrics to them so they
are not preferred over the routes learned by the provider "B" router.

Is this possible or would I need to be running different EIGRP
processes between the two customer prem routers and then play around
with some redistribution? I am hoping this isn't the case because I
don't have access to those CPE routers and redistribution is a nasty
thing...

Thanks in advance for any enlightnment.

Cheers,
Mike





More information about the NANOG mailing list