When EIGRP removes a successor route (a route learned from a successor – or in other words the next hop for this route) and no alternative path exists, meaning there is no feasible successor in the topology table ( show ip eigrp topology ) the router starts the process of finding a new loop-free topology. Then the state of the DUAL algorithm goes to active. Using the show ip eigrp topology command, the route is marked with “A”.
e.g.
show ip eigrp topology
A 10.0.0.0/30, 0 successors, FD is 2172416
A query message is send to neighbors. Until a reply is received, the router cannot use any alternative paths for this route. In the half of the active timer (by default 3 minutes) the router sends a SIA-Query to notify them that yet the router has not received any reply. If the router gets a SIA-Reply means that it waits also for other routers to reply back. However if no answer is received the router send up to 3 more SIA-Query packets in half-active-time interval. If no answer is received, the router brings down the adjacency with the routers that haven’t replied to the query.
Should the adjacency come back when Hello packets are received?
The Hello messages indicate that the EIGRP runs on the router. However, in case that the route is Stuck-in-active, logs:
—-R1—–
DUAL-5-NBRCHANGE: IP-EIGRP(0) 1: Neighbor 192.168.2.7 (FastEthernet1/0) is down: holding time expired
—-R2—–
DUAL-5-NBRCHANGE: EIGRP-IPv4 1: Neighbor 192.168.2.8 (FastEthernet0/1) is down: retry limit exceeded
!
DUAL-5-NBRCHANGE: EIGRP-IPv4 1: Neighbor 192.168.2.8 (FastEthernet0/1) is up: new adjacency
and the adjacency is not coming back. In order to form the adjacency again (at least in my lab) the interface should be shut down first.
Some references:
http://www.cisco.com/c/en/us/support/docs/ip/enhanced-interior-gateway-routing-protocol-eigrp/13676-18.html
https://tools.ietf.org/html/draft-savage-eigrp-00#section-5.3.2
