[c-nsp] EIGRP feasible successors

[Bruce Pinsky]

On 10/3/2014 7:03 AM, Howard, Christopher wrote:
> I'm hoping for some clarification as to whether I'm incorrect or my switch
> is incorrect.
> 
> I have a switch (4500X) that has 3 different routes to another switch.
> Two routes traverse 10G links and the other is a 1G link.  However,
> traffic is getting transferred through the 1G link thanks to EIGRP.  I
> think EIGRP is wrong.
> 
> 
> First, the topology table says it has 3 successors, but only lists 2.  I
> have filtered out to just one subnet, but there are others this way.
> 
> switch#sh ip eigrp vrf green topology
> P 172.1.2.0/24, 3 successors, FD is 3072
>         via 10.1.1.6 (3072/2816), Vlan910
>         via 10.1.9.6 (3072/2816), Vlan2910
> 
> 
> If I tell it to show me all links in the topology table, I can see the
> third route.
> 
> switch#sh ip eigrp vrf green topology all-links
> P 172.1.2.0/24, 3 successors, FD is 3072, serno 25436
>         via 10.1.1.6 (3072/2816), Vlan910
>         via 10.1.9.6 (3072/2816), Vlan2910
>         via 10.1.5.3 (3328/3072), Vlan1910
> 
> 
> Now, as I understand it, the first two routes are successors because they
> have the lowest feasible distance.  The third route should not be
> considered a feasible successor because the advertised distance is equal
> to the feasible distance of the successors (the feasibility condition
> explicitly states less than).  However, it appears that the switch is
> considering this third route as a successor.
> 
> 
> And worse, due to the use of the variance command, the switch is using the
> third route as the active one.
> 
> switch#sh ip route vrf green 172.1.2.0
>   Last update from 10.1.9.6 on Vlan2910, 3w5d ago
>     10.1.9.6, from 10.1.9.6, 3w5d ago, via Vlan2910
>       Route metric is 3072, traffic share count is 40
>   * 10.1.5.3, from 10.1.5.3, 3w5d ago, via Vlan1910
>       Route metric is 3328, traffic share count is 37
>     10.1.1.6, from 10.1.1.6, 3w5d ago, via Vlan910
>       Route metric is 3072, traffic share count is 40
> 
> 
> I can remove the variance from EIGRP so that route will drop from the
> route table, but am I incorrect in thinking that route should not be a
> feasible successor in the first place?
> 
> 

No, there are two conditions that must be met for variance to work.  The
first is this one:

"The route must be loop-free. This condition is satisfied when the reported
distance is less than the total distance or when the route is a feasible
successor."

I think you are getting bit by the first part of the condition because the
computed total distance is 3328 and the reported distance is 3072.

Now variance works as multiples of the metric of the best route.  So, if
you have any variance other than 1, that third path is going to be installed.

-- 
=========
bep