2

I am trying to configure unequal cost load balancing in EIGRP and am a little confused about a couple of parts. I have an ipsec gre tunnel using comcast, as well as a point to point t1 that I want to balance across.

I have set the variance, set traffic-share to balanced, enabled IP CEF and on each interface set ip load-sharing per-packet.

My routing table shows EIGRP routes across both the tunnel interface as well as the T1 to the networks I am interested in.

Do I need to restart the interfaces on the router to have the ip load-sharing per-packet command take effect? Are there any other steps that I need to take to make this function correctly?

user27787
  • 113
  • 8

2 Answers2

0

I wouldn't think so -- if the information is in the routing table it should be in affect immediately.

One word of warning though: you probably want to use 'per flow' load-balancing and not 'per packet' otherwise you are going to have some horrible performance issues.

Peter
  • 5,403
  • 1
  • 25
  • 32
0

I think Peter has it right. It sounds like it's already working if the info is in your routing table already. Try using 

show ip eigrp topology

It will give you a list of all feasible successors, if both are in your list it should be load balancing.

smoak
  • 646
  • 2
  • 7
  • 13
  • show ip eigrp topology shows 1 successor for each route, so they all basically look like this: P 10.10.11.0/30, 1 successors, FD is 2693120 via 10.128.5.5 (2693120/2181120), Serial0/0/0 via 10.128.6.1 (285214976/2181120), Tunnel1 Show ip route looks like this for each route D 10.10.11.0/30 [90/285214976] via 10.128.6.1, 1d08h, Tunnel1 [90/2693120] via 10.128.5.5, 1d08h, Serial0/0/0 If I do a show int Tunnel1 the 5 minute input/output counters are 0, but on s0/0/0 it is running the T1 at around 94% all the time... – user27787 Dec 03 '09 at 05:21
  • 1
    From your response it looks like it's set up correctly since it's using your T1 about 94% of the time. If you use show ip route 10.10.11.0 it will show you some more statistics (look for traffic share count). In order to get a better ratio you may have to increase the delay on your local serial interface (which would increase the metric). See http://www.ccietalk.com/2008/09/21/eigrp-unequal-cost-load-balancing for a bit more information. – smoak Dec 03 '09 at 23:20