EIGRP route selection weirdness

Unanswered Question
May 23rd, 2008

In an EIGRP setup in my study lab, I have routers G, H, and K all connected via TokenRing on subnet

Router K (a 2513) also has an Ethernet connection, to Router A, on subnet

Back in the TokenRing setup, G has a 9600bps serial link to Router J ( And finally, J and A have a 9600bps serial link to each other, (Each interface address ends with the number equaling its place in the alphabet, i.e A's interfaces' last octets are always .1; J's are always .10, G's are always .7 etc.

A snippet of K's routing table reads: is subnetted, 2 subnets

C is directly connected, TokenRing0

C is directly connected, Ethernet0 is variably subnetted, 3 subnets, 2 masks

D [90/2233856] via, 00:55:25, TokenRing0

D [90/2233856] via, 00:55:25, TokenRing0

D [90/23310336] via, 00:55:31, Ethernet0 is variably subnetted, 2 subnets, 2 masks

D [90/2195456] via, 01:00:37, Ethernet0

D [90/2195456] via, 01:00:37, Ethernet0

A thing I can't figure out is why Router G's serial interface ( is appearing in K's routing table with A as its successor. K is one 4Mbps hop away from G; to get there via A it has to go over the Ethernet link (10Mbps, sure SOUNDS good) but then over a slow serial link to J and then yet another slow serial link to G. I haven't configured any eigrp variance either; so I might expect to see this sort of thing in the eigrp topology table, but surely not the routing table. What am I failing to understand here?

  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 0 (0 ratings)
michael.leblanc Fri, 05/23/2008 - 18:13

Perhaps you should take a look at the "bandwidth" setting on each of the interfaces (particularly the serial interfaces). This value is metric relevant.

router#sh run int e0

Building configuration...

Current configuration : 415 bytes


interface Ethernet0

bandwidth 5000

CriscoSystems Fri, 05/23/2008 - 18:33

Which, I just did. Eth0's BW is 10000 kbps. Router A's Ser1 (the link to J) is 1544; and J's Ser4 (link to Router G) is actually BW 115 kbps.

While K's token ring interface is bw 4000 kbps.

With a 115 kbps link in there, this seems to make even less sense.

royalblues Sat, 05/24/2008 - 02:41

EIGRP does not consider cumulative bandwidth to the destination

It only considers the lowest bandwidth and only the delay is cumulative which in your case is 9600kbps via both the paths

So if the FD is less via A, it will choose it as a successor



michael.leblanc Sat, 05/24/2008 - 10:33

Perhaps "compare" the BW & Delay settings on the interfaces in "both" paths.

A table of BW & Delay of each interface on each device would be helpful.

Maybe do a "sh ip protocol" on each device and confirm that the metric weights are consistent on all devices.

e.g.: EIGRP metric weight K1=1, K2=0, K3=1, K4=0, K5=0

michael.leblanc Mon, 05/26/2008 - 10:22

Might want to confirm that Router G is using the same EIGRP Process ID as the others, and that you don't have a missing/incorrect network statement for G's serial interface.


This Discussion