Two EIGRP routes, prefer one over the other

[TRACY HARTMANN]

We are running EIGRP and we have one site that has the following:

Small point to point and a large bandwidth through MPLS.   The issue is I have a assymetric route because the site perfers the MPLS since that is the default gateway but the return traffic prefers the local point to points since there route is better.   How can I make the return traffic go across the MPLS.  I have tried to add a distance command but not sure I did it right.

Thanks

[acampbell]

Hi,

You could try changing the bandwidth statements to make sure the point to point is less than the MPLS
for their relevant interfaces.

Also if you change the ADmin Distance you should issue clear ip route to invoke new value.

See this link from TAC

https://www.cisco.com/c/en/us/support/docs/ip/enhanced-interior-gateway-routing-protocol-eigrp/13673-14.html
 

[Richard Burts]

If you are trying to manipulate the metric for an EIGRP route it is generally better to change the delay rather than the bandwidth.

I am not clear about the topology involved in this issue. Do both sites have the MPLS connection and the point to point connection? Perhaps the original poster can post the output of show ip route from both routers pointing out which networks in the output are the ones involved in this issue?

HTH

Rick

[TRACY HARTMANN]

Sorry if it wasn't clear.

There is a host end that has a MPLS connection and 3 meg point to point.   Two difference routers, one managed by us and one by the vendor.

At the remote end is the same thing a MPLS connection and 3 meg point to point.

The remote end works fine because the default gateway is the MPLS router so it routes outgoing just fine.  The return route however goes over the point to point since it is a better.  I am assuming I should put a delay on the Host end point to point router?

network:

Remote p2p router -----layer 2 switch ---MPLS router

Host psp router ------layer 3 core switch ---- MPLS router

This is the Host end.

Routing entry for 10.4.64.0/24
  Known via "eigrp 800", distance 90, metric 3395840, type internal
 

Routing Descriptor Blocks:
  * 10.4.150.154, from 10.4.150.154, 3d22h ago, via Multilink6
      Route metric is 3395840, traffic share count is 1
      Total delay is 100100 microseconds, minimum bandwidth is 3072 Kbit
      Reliability 255/255, minimum MTU 1500 bytes
      Loading 170/255, Hops 1

[Jon Marshall]

It depends and it is still not clear. 

At the host end are you redistributing BGP from MPLS into EIGRP or are you runnng EIGRP over MPLS.

If you are using BGP and redistributing then the L3 switch will see the P2P EIGRP route as AD 90 whereas the MPLS EIGRP route will be EIGRP 170 because it has been redistributed in which case delay will have no effect at all. 

We really need more details about the above. 

Jon

[cofee]

To my knowledge eigrp routes should be learned as internal eigrp with an AD of 90 over the MPLS cloud through IBGP vpnv4 as long as eigrp austonomous system is same on both PE routers because all the eigrp attributes will be sent over to the other side of tunnel via extended community.

I would go with manipulating delay on the point-to-point links between sites to prefer MPLS links for inbound and outbound traffic. Advertising prefixes with a smaller match using leak map over point-to point link can also work but that would depend on how many prefixes are there and if they can be aggregated or not, but that may complicate the over all configuration.

@lespejelalso made a good suggestion of using the access list to increase AD for prefixes learned point-to-point link.

[lespejel]

hello

@Richard Burts is right, you should influence routing by increasing delay and turning that interface less desirable; second step is to do the same on the other site to avoid asymmetric routing.

If you have 2 possible paths, the secondary path should have an increasing delay on bot sites.

If you want to balance traffic, meaning, some routes prefered over link A and some others over link B, then you can use a distance command with an ACL filtering which routes to change the AD.

example:

distance 95 1.1.1.1 0.0.0.0 99 

distance AD, [neighbor wildcard] [ACL]

regards.