This situation is something that happens on occasion to a wide-area network running EIGRP. It tends to freak people out when they see it if they're not very familiar with this Cisco proprietary protocol, and they anticipate a meltdown. However, it's not a cause for concern unless it happens consistently or for a long period of time. What it comes down to is an outstanding route query, somewhere. Your job - if that is your job; if not, why are you reading this - is to track the cause of the oustanding query. It may be another router with a problem, or a number of other things (see end of w-u). Regardless, all you can do is follow the routes while using ping until you either find something interesting or go completely insane.

More to the point, a route is active if it's lost, and if it doesn't have a successor or a feasible successor. Stuck-In-Active routes indicate two problems: that route/s are active and stuck. Both 'Why is this route going active' and 'How is this route getting stuck' need to be answered, though the latter cause is more important to find. If the hold-down timer, which is three minutes by default, expires before all queries are answered, the router reinitializes neighbor relationships, goes active on all routes known through the unanswering neighbors, and readvertises all routes to the bounced neighbor.

When a route goes active, the router sends out queries to all neighbors (though not through the interface where the route was lost), and waits for replies. Say this first router is dannye. When a route on dannye goes active, a 3-minute timer begins. The timer will run until until a reply is sent back from the next router downstream. But the second router, knifegirl, has been querying other routers and has not been able to find a feasible successor. Knifegirl has aso gone active. If dannye's timer has expired, he will consider the path through knifegirl unreliable. Dannye will reset his neighbor relationship wiith knifegirl, and relearning those routes requires rebuilding the neighbor relationship.

Using the DUAL-SIA error messages, you can see if the routes that are going active are consistent, if there is a commonality; e.g., are they all /27s, or from the same network region? There may not be a commonality, and this makes troubleshooting more difficult. If the route is staying active for more than a minute, then there's a problem. Look for lossy and loaded links and misconfigured bandwidth parameters.

Say that dannye has been waiting on knifegirl for 1:42. What is her deal? Go to knifegirl and look at the results of show ip eigrp top active (if the problem is occurring right then) to see why dannye hasn't received a reply from her. It looks like knifegirl is still waiting on a reply from hatless, one router down. Can't these noders get it together?

Do a show ip eigrp top active on hatless. He's in the same condition. Go to MoJoe, one router down. He's not waiting on any other router, which indicates that the link between hatless and MoJoe is unreliable. Go back to hatless and do some pings, see if everything's kosher. You basically have to chase down the active routes. Maybe you can't narrow it down to a bad link between two routers, like hatless and MoJoe. Maybe it just keeps going. You could quit your job and go find the mythic nitrogen pudding baths on Planet Heineken, or try to do some more summarizing. Your call.


The active routes may be a result of memory shortage, congestion on the link, a bad link, an MTU problem (e.g., large packets aren't being delivered), excessive redundancy, or an excessively long query range.


Basically. :-)