In addition to calculating the successor (best next hop), DUAL will also attempt to find one or more “feasible successors” (loop-free backup next hops) for each destination. Refer to Figure 1 for an example:
We’ll be looking at things from R1’s perspective. As previously discussed, because the Feasible Distance (FD) via R2 (2000) is less than the FD via R3 (2500) or R4 (3300), R1 will choose R2 as the successor for the 10.1.1.0/24 subnet. Next, DUAL will attempt to find feasible successors, which are loop-free backup next hops for the destination. In the example, both R3 and R4 provide alternate paths to the destination. In order to qualify as a feasible successor, the candidate must meet the “feasibility condition”, which is that the Advertised Distance (AD) received from the candidate must be less than the FD via the current successor. This feasibility condition ensures that the candidate for feasible successor isn’t dependent on the current successor for connectivity to the destination (if it were, the result would be a routing loop). Note that the condition specifies “less than”, and not “less than or equal to”, which takes care of the case in which the link metrics are allowed to be zero, as they are with some routing protocols.
Since the AD received from R3 (1500) is less than the FD via the current successor (2000 via R2), R3 does meet the feasibility condition, and would be flagged as a feasible successor for 10.1.1.0/24 in the EIGRP topology table. What about the path via R4? Because the AD of R4 (2300) is not less than the FD through the current successor (2000, via R2), R4 does not qualify as a feasible successor for 10.1.1.0/24. So, the current situation is that R2 is the successor for 10.1.1.0/24, and R3 is the feasible successor.
If R1 detects a problem with the current successor, it would immediately promote the feasible successor to successor. In this case, R3 would become the successor for 10.1.1.0/24, and R1 would attempt to find a new feasible successor. Because R4’s AD of 2300 is less than the FD via the current successor (2500 via R3), R4 does qualify as a feasible successor for 10.1.1.0/24, and it would be flagged as such in the EIGRP topology table. If the path via R3 was now lost, R4 would be promoted to successor, and there would be no feasible successor. This pre-calculation of a backup path is unique to EIGRP, and it is the reason that EIGRP converges faster than the other IP routing protocols.
Next time, we’ll look at situations in which no feasible successors exist.