Feroz, Alain,
Please be careful with the term "Feasible Distance" - it is not the current distance through a particular neighbor. The Feasible Distance to a destination is the smallest known distance to the destination since the last time the route transitioned from Active to Passive state. The Feasible Distance is computed per destination (i.e. one FD per destination network), not per neighbor. Also, being the smallest known distance since the last Active->Passive transition, it may be different (lower) than the current distance to the destination.
The resulting distances to a destination if going via a particular neighbor do not have an official name in EIGRP terminology. Don Slice once mentioned that internally in Cisco, they call them Computed Distance. Personally, I would call them Total Distance because that is what they are about - total distances to the destination over a particular neighbor, computed as the sum of the Reported Distance of that neighbor, plus the link cost towards that neighbor.
To reuse the example above, this would be the correctly used terminology:
30 / 10 TD/RD feasible successor
20 / 10 TD/RD successor, 20 is the lowest distance so let's also take it as the FD
45 / 25 TD/RD
Because here, we have set FD to 20 (assuming this is the initial startup of the network), the first and second neighbor pass the feasibility condition, the third neighbor does not pass it. Now, let's change the reported distances somewhat, keeping the link costs intact:
25 / 5 TD/RD feasible successor
22 / 12 TD/RD successor
45 / 25 TD/RD
Now, the total distances have changed. What sequence of steps has happened here?
- When the first and second neighbors reported a changed RD, their RD still met the feasibility condition using the FD of 20. 5 is less than 20 (first neighbor), 12 is less then 20 (second neighbor). Therefore, they still meet the feasibility condition.
- The smallest total distance path after the RD update goes through the second neighbor, and as we explained in the previous step, the second neighbor still meets the feasibility condition, therefore we do not need to enter the Active state. The second neighbor will therefore become the successor (or better said, it will retain that role), and the route will remain Passive throughout the whole update.
- However, the new current smallest total distance is now 22, while the smallest known distance to the destination stored in the FD is currently 20. Because the current smallest distance is higher than the FD, and the FD is defined as the minimal known distance since the last Active->Passive transition, the FD will not be updated and will stay set to 20. The router will have the route in its routing table with the metric of 22 and it will advertise this metric of 22 to its neighbors, but internally, for feasibility checks, it will always use the value of FD, in this case, 20.
Now, in this changed metric scenario, if variance 3 was configured, the router would be willing to put worse routes to the destination into its routing table if they meet the following two criteria:
- They must go over a feasible successor, and
- Their Total Distance must be at most 3 times higher than the current smallest metric
The first requirement limits our choice of paths only to the first and second neighbor. Because the third neighbor does not meet the feasibility condition using the FD of 20, it will not be considered for unequal-cost load balancing.
The second requirement will now allow the first neighbor to be used because our current smallest metric is 22, so the with the variance 3, any route through a feasible successor whose total metric is at most 22*3=66 is accepted. The route through the first neighbor has a total metric of 25, so it is well within the allowable range.
Please feel welcome to ask further!
Best regards,
Peter
