Nexus 5596U EIGRP route preference

[Rahsaan White]

Hello.

I have a customer with a primary datacenter and DR datacenter, that has a P2P 100Mbps link connection between them.  At the primary Datacenter there will be a Nexus 5596U with a L3 card running EIGRP, it will have three connections, one nothbound to the Internet not a problem, the other is north bound to an MPLS SP managed ISR router.  Both datacenters will have MPLS SP managed ISR router, the MPLS service provider will be redistributing BGP into EIGRP from their ISR routers at both datacenters.  This means we will learn external EIGRP routes with an AD of 170.

Connected to my 5596 southbound will be the SAN for the EMC storage traffic, the DR also has a SAN with EMC storage as well.  The 100Mbps P2P link is primarily for replication traffic.

Since the Nexus 5596U with L3 module doesn't support PBR I have to figure a way for replication traffic to prefer the 100Mbps P2P link vs the MPLS cloud.  I was thinking of running iBGP over the P2P link with the Nexus 5596 being a route reflector and it's iBGP peer which will be a Catalyst 3750X at the other end of the P2P link being the route-reflector client.  That way both iBGP peers will exchanged learned routes.

I have to come up with a way for the replication traffic that comes sourced from the IP addresses of the EMC/storage to prefer the P2P link which will have an AD of 200 from iBGP in the routing table vs the external EIGRP traffic that has an AD of 170 in the routing table.  All other traffic will route normally across the MPLS cloud.  All other traffic will include client server traffic as there are application servers that sit south bound of the Nexus 5596, the client traffic will come inbound to the datacenter via the MPLS cloud.

I don't have a Nexus 5596 to do a mock lab.  My thought is to redistributed connected with a route-map in EIGRP on the Nexus 5596 with a two match statements then set the IP next hop to the IP of the 3750X.  I would need to select only replication traffic, so I was thinking if I can match on vlan and next-hop then my set command would be the next-hop of the 3750X. 

Any suggestions are appreciated.

[Kelvin Willacey]

If I understand correctly you have two links between the DC and the DR, the MPLS and the point to point 100Mbps link and you would like the replication traffic to prefer the 100Mbps link. If that is the case you should not need to run iBGP since, and I am assuming here, the point to point link is a direct connection between the DC and DR and if you were to run another EIGRP AS/process over that link then all the routes learned via this process would be internal with an AD of 90.

If the 100Mbps link is only for replication then I would also suggest only including the subnets for the storage network and the point to point link in the EIGRP AS/routing process that way no other traffic can flow over this link except the replication traffic.

If the routes learnt via both links have an AD of 170, you can also use an off set list to influence the metric of specific routes via a particular link.

[ugot2nome]

Your solution is correct.

Just a clarification, as long as the EIGRP AS numbers match at both DCs, and the SP is sending extended communities within their MPLS cloud; the BGP process during redistribution into EIGRP does a flag check and if the AS match; the route will be put into EIGRP as internal instead of external.