EIGRP OTP (Over the Top) allows you to run EIGRP between routers that are not directly connected. It is a great feature to use when you want to run EIGRP between routers that are connected to a service provider network and you don’t want the hassle of other solutions like MPLS VPN and you don’t want the service provider’s involvement with your routing. With EIGRP OTP, you can configure everything on your own routers as long as you are able to reach the remote router(s).
EIGRP OTP uses an overlay VPN and thus uses tunneling, it is a bit similar to other solutions like DMVPN / multipoint GRE but it uses LISP (Locator/Identifier Separation Protocol) instead.
LISP is a technology where each end device has a separate “identity” and “location”, unlike IP where the IP address is both the identity and location of the end device. LISP is useful in scenarios where an end device should be able to retain its original IP address, even if it moves to another subnet. It is useful for VM motion where virtual machines are moved from one hypervisor to another but also for mobile devices that end up in another subnet after roaming.
There is a control plane and data plane for LISP. Learning all the details about the LISP control plane is quite a story. On the data plane, LISP uses UDP to encapsulate data and tunnel traffic. EIGRP OTP only uses LISP for the data plane, EIGRP is used for the control plane. This is convenient as you don’t really have to understand LISP to be able to understand EIGRP OTP.
In this lesson, I’ll show you how to configure EIGRP OTP.
To demonstrate EIGRP OTP, I use the following topology:
Above we have four routers:
- R1, R2 and R3 are customer routers that want to use EIGRP to exchange routing information. These routers each have a loopback interface with an IP address that we will advertise in EIGRP.
- R4 is the service provider network. In a production network, this could be a large network but for simplicity reasons, I’m only using one router here.
The only requirement to run EIGRP OTP is that the customer routers have to be able to reach each other. I will use a couple of static routes so that R1, R2 and R3 are able to reach the IP addresses on their GigabitEthernet 0/1 interfaces:
R1(config)#ip route 192.168.24.2 255.255.255.255 192.168.14.4 R1(config)#ip route 192.168.34.3 255.255.255.255 192.168.14.4
R2(config)#ip route 192.168.14.1 255.255.255.255 192.168.24.4 R2(config)#ip route 192.168.34.3 255.255.255.255 192.168.24.4
R3(config)#ip route 192.168.14.1 255.255.255.255 192.168.34.4 R3(config)#ip route 192.168.24.2 255.255.255.255 192.168.34.4
Now we can configure EIGRP.
EIGRP OTP is only supported in EIGRP named mode, you can’t configure it in classic mode. Neighbors have to be configured statically and we have two options here:
- We can configure a full-mesh of static EIGRP neighbors.
- We can configure one router as a route reflector which works similar to the BGP route reflector.
With three routers, a full mesh is no problem but if you have a lot of customer routers, a route reflector is an interesting option. I will configure R1 as the route reflector and R2/R3 will be its clients:
R1(config)#router eigrp OTP R1(config-router)#address-family ipv4 autonomous-system 123 R1(config-router-af)#remote-neighbors source GigabitEthernet 0/1 unicast-listen lisp-encap 123
The remote-neighbors command is all we need for EIGRP OTP. It tells R1 to use the GigabitEthernet 0/1 interface and listen for unicast packets. LISP uses different IDs, I’m going to use ID 123 on all routers. Optionally, this command allows you to set a maximum number of neighbors and/or use an access-list to permit only certain IP addresses to accept as EIGRP neighbors.
Let’s advertise the loopback interface and GigabitEthernet 0/1 on R1:
R1(config-router-af)#network 184.108.40.206 0.0.0.0 R1(config-router-af)#network 192.168.14.0 0.0.0.255
Last but not least, there are two more things we have to configure:
- Disable next-hop-self: If you don’t do this, R1 will set its own IP address as the next hop for everything it advertises to R2 and R3. If you don’t disable this, then whenever R2 or R3 want to reach each other’s loopback interface, they will send their traffic to R1. It works but it’s sub-optimal routing. If you disable this then R1 will leave the next hop IP address alone and R2/R3 are able to reach each other directly.
- Disable split horizon: R1 is using a single interface so if you don’t disable split horizon, R2 and R3 are unable to learn each others loopback interfaces.
Let’s disable next-hop-self and split horizon:
R1(config-router-af)#af-interface GigabitEthernet 0/1 R1(config-router-af-interface)#no next-hop-self R1(config-router-af-interface)#no split-horizon
That’s all we have to do on the route reflector.
Let’s configure R2 and R3. The only thing we have to do is configure R1 as a static remote neighbor and set the LISP encapsulation ID: