EIGRP and Metro E

Unanswered Question
Sep 23rd, 2008

I seem to have an issue with EIGRP as it relates to detecting an outage. During some longer outages when we lose layer 3 connectivity between sites, EIGRP does not bounce. I am assuming this is because the physical interface does not bounce. How dynamic routing protocols should be configured on Metro E to detect outages more seamlessly?

I have this problem too.
0 votes
  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 0 (0 ratings)
John Blakley Tue, 09/23/2008 - 08:21

I would suggest adding a static route to your next hop in case of an outage:

ip route preferredIP

ip route backup 254

The 254 route won't get used unless the preferred doesn't respond.


Joseph W. Doherty Tue, 09/23/2008 - 08:34

When the interface remains up, but the peer is down, many routing protocols rely on some keepalive method to detect loss of peer or neighbor. Default peer loss detection periods can be rather long but they often can be reconfigured to decrease the time.

For EIGRP, you could adjust the hello interval timer and the hold timer. (The default for you might be an interval of 5 seconds and 3x for hold period, i.e. 15 seconds to detect loss of peer or neighbor.)

plavine Tue, 09/23/2008 - 09:45

That seems like it should work on paper or in a lab however in production I will sustain losses of minutes before I finally passive out the interface to force eigrp to converge. It is almost as though the keepalive timers are not working at all.

Joseph W. Doherty Tue, 09/23/2008 - 11:44

So, you've already set the hello timer down to 1 second with the default 3x? In other words, neighbor loss isn't detected in about 3 seconds?

If the problem is 3 seconds still isn't fast enough, some of the newer platforms with a later IOS support BFD with EIGRP; i.e. subsecond neighbor loss should be detected. For more information about that see:



h.parsons Tue, 09/23/2008 - 12:35

I don't think adjusting timers will fix your problem, if it is taking minutes by using the default timer settings then you have problems elsewhere.


This Discussion