EIGRP Unequal Load sharing

[MARK GRABER]

I'm having trouble getting EIGRP load sharing to work. I have 4 6509s, 2 each at locations in NY and CA. Each 6509 has a FlexWan card with a CT1 interface, connected via T1 Frame-Relay to the opposite side of the country like this:

CAL NY

6509 --T1-- 6509

| |

GigE GigE

| |

6509 --T1-- 6509

I've set the variance to 2, but I'm not getting load balancing; I'm seeing the second routes in the EIGRP topology table, but they're not making it into routing table. Here's a sample of the Topology database:

IP-EIGRP topology entry for 10.33.15.0/24

State is Passive, Query origin flag is 1, 1 Successor(s), FD is 2178816

Routing Descriptor Blocks:

10.33.3.1 (Serial4/0/1:0.200), from 10.33.3.1, Send flag is 0x0

Composite metric is (2178816/2816), Route is Internal

Vector metric:

Minimum bandwidth is 1536 Kbit

Total delay is 20010 microseconds

Reliability is 255/255

Load is 6/255

Minimum MTU is 1500

Hop count is 1

10.32.0.11 (Vlan3), from 10.32.0.11, Send flag is 0x0

Composite metric is (2179072/2178816), Route is Internal

Vector metric:

Minimum bandwidth is 1536 Kbit

Total delay is 20020 microseconds

Reliability is 255/255

Load is 151/255

Minimum MTU is 1500

Hop count is 2

Any suggestions?

[Richard Burts]

Mark

The real problem here is that the second path does not pass the feasibility test. Even though the calculated metric for both paths winds up being fairly close (2178816 for the first path and 2179072 for the second path) the advertised distance for the second path of 2178816 is equal to the local metric of the first path. As such EIGRP does not trust it to be loop free and will not put it into the routing table - no matter what you do with variance.

If you want it to work, you may be able to do something like an offset list that will increase the advertised metric of the first path enough to allow the second path to pass the feasibility test.

HTH

Rick