EIGRP - Delay Convergence

Wantser1981_2 · ‎06-14-2010

Hi,

Does anyone know of a way of stopping EIGRP from converging back to a link if it is more preferable once it becomes available again?

Scenario:

Two sites, two links, 100Mb primary and 10Mb back up. EIGRP domain across the two sites, but the 10Mb is BGP which is redistributed.

I have experienced an issue with a flapping link where the timers expire, link goes down, then comes back up for 30 seconds, converges back and then repeates 111 times!

To reduce the times this happens, am I able to hold the route on the resilient link for a period to give the primary link chance to settle, or have EIGRP "test" the preferable link for a time before it is chosen?

The latter I dont think will be possible, but is there a hold timer that can be configured for the first....?

Thanks in advance

andrew.prince · ‎06-14-2010

EIGRP timers by default - are 5 second hello and 15 seconds hold. Typically the hold is 3 times the hello - these timers are configurable on a per link basis.

On the 10mbs link interface you can configure the timers to be longer or shorter - whatever you prefer, the config is:-

ip hello-interval eigrp <> <>
ip hold-time eigrp <> <>

HTH>

Wantser1981_2 · ‎06-15-2010

Hi Andrew,

Thanks for the info.

I am okay with the hello and hold timers of the neighbours, however am interested in trying to work out how I might hold on to a path in the active topology when a relationship is formed and a more desirable route is identified. So the oposite to loosing a neighbour. A wait time when learning about one so that I can reduce the risk of multiple topology changes when a more desirable link is flapping 3-4 times a minute and traffic is stuttering due to the hold times of loosing a neighbour. I dont want to decrease these times preferably.

Cheers

andrew.prince · ‎06-15-2010

The whole point of a dynamic link state/enchanced distance vector routing protocol - is fast link fail detection and fast convergance!

The you should think about using another dynamic routing protocol - perhaps RIP v2.

Wantser1981_2 · ‎06-15-2010

I get the whole point of a dynamic routing protocol thanks, but I am concerned with the topologies return to normal after the most attractive route fails.

But what I am asking, is is there anyway to protect against a better primary link that is constantly failing in the frequency I have mentioned which means the quick convergency, that is required on total failure (off, that is it!), doesnt become an issue if the link is bouncing around all over the place causing quick convergence after quick convergence. A back off period, if you like, which means that instead of installing a better link straight away, a time can be chosen to wait so that you increase network stability even if it is for 10 minutes for example. Stay on the backup link for a period, before going back to primary when it become available. Similar to STP Listening and Learning I suppose.....

Like I said, this may be wishful thinking as I dont recall such a feature....hence the question.

Thanks

andrew.prince · ‎06-15-2010

Ahhh OK I get it now - sorry for being dense!

Yes - EIGRP has a feature "metric holddown" not sure if it's configurable though on the amount of time to wait etc.

HTH>

Richard Burts · ‎06-15-2010

I wonder if the original poster has thought about using Interface Dampening. It seems to me that this feature would address the issue of an attractive interface coming back up but flapping.

This link has some information that could get you started:

http://www.cisco.com/en/US/docs/ios/12_3t/ip_route/command/reference/ip2_c1gt.html#wp1093971

HTH

Rick

HTH

Rick

andrew.prince · ‎06-15-2010

Hi Rick,

I did think that, but then using VRF's always does my head in when I'm not doing MPLS!!!

But then I should read before I reply.....did not know of the specific interface command - v.nice.

Message was edited by: Andrew Prince