Arista VXLAN 1 – VXLAN, HER and EVPN Basic Configuration

This lab use part of the Arista Spine and Leaf topology I have created in GNS3. Specifically, I will only be using three switches. One spine and two leaf switches. I have chosen this to get a good understanding of the core concepts of VXLAN, HER and EVPN.

This is the basic VXLAN topology that I will use to demonstrate the VXLAN configuration and how to get the hosts to communicate. I will be using two methods;
– Head End Replication
– EVPN

Head End Replication Or EVPN?

There are differences between VXLAN Head End Replication and EVPN (Ethernet Virtual Private Network). Mainly flooding on the part of HER. I like to think of the difference between the two like multicast sparse mode and dense mode.

EVPN (Ethernet Virtual Private Network) used MAC and ARP learning that allows EVPN to provide a distributed control plane from MAC and ARP learning across the DC fabric. It allows each VTEP to only learn about the MAC addresses and ARP entries associated with specific VXLAN segments.
It avoids unnecessary flooding of BUM traffic to all VTEPs by leveraging the MAC and ARP information learned through the control plane.

This is different to how Head End Replication works which uses flood and learn behaviour. When a VTEP receives a packet with an unknown destination MAC address, it floods the packet to all other VTEPs in the overlay network. The destination VTEP then learns the location of the MAC address and installs an entry in its forwarding table. This may causes scalability issues as a topology grows larger

Configuration

For this lab I have used vEOS 4.26.2F
Device configuration will be in stages to confirm each step as we go. For the complete device configs, go here. And for individual config section, see links below.

Underlay Configuration
VTEP Configuration
VXLAN Head End Replication Layer 2
VXLAN Head End Replication Layer 3
VXLAN EVPN – No VRF
VXLAN EVPN – With VRF
VRF Route Leaking

IP Addressing Scheme

SwitchEth1 (eBGP UL)Eth2 (eBGP UL)Lo0 (BGP EVPN)Lo1 (VTEP)VLAN11VLAN12VLAN20VLAN30
Spine110.1.11.0/3110.1.12.0/3110.10.10.1/32
Leaf110.1.11.1/3110.2.11.1/31
(Spine 2)
10.10.10.11/32172.20.1.1/3210.10.11.254/24192.168.12.254/24192.168.20.1/24192.168.30.1/24
Leaf210.1.12.1/3110.2.12.1/31
(Spine 2)
10.10.10.12/32172.20.2.2/3210.10.11.254/24192.168.12.254/24192.168.20.1/24192.168.30.1/24

Underlay

The underlay is using eBGP. There are several different designs for the underlay, which I am not going to get into here. My understanding is that eBGP is the most flexible and widest used. I have a discussed other options in a design guide.

Spine 1
Leaf 1
Leaf 2
Testing Commands

VTEP

Leaf 1
Leaf 2

VXLAN Head End Replication (HER) Layer 2

There is a template to follow the VXLAN config that is below, this is specifically for HER – Head End Replication.

Leaf 1
Leaf 2
Testing

This is for VLAN 12 only and therefore is layer 2 only. No SVIs are required.

ARP Over VXLAN
ICMP Over VXLAN

VXLAN Head End Replication (HER) Layer 3

Leaf 1
Leaf 2
Testing

Traffic from VLAN 11 to VLAN 12 going via Spine 1 VXLAN encapsulated from Leaf 1 to Leaf 2 and also interVLAN routing on Leaf 1.

Ping between two networks over VXLAN
Ping between VLANs on Leaf 1, no VXLAN

VXLAN EVPN No VRF

Any HER configuration needs to be removed on the VXLAN interface. EVPN must also be enabled using a service command.

This example creates a two VLANs 11 and 12 in the same way as HER did, but this time using EVPN. I had a few issues with this config.
– I needed to redistribute connected into Spine 1 BGP
– I did not use the command no bgp default ipv4-unicast on the leaf switches
Both of these commands would cause the BGP neighbourships to fail.

Spine 1
Leaf 1
Leaf 2
Testing

BGP EVPN Capture shows the BGP networks that are shared, as well as the fact that EVPN is being used.

Leaf 1 BGP Update Message

Each leaf switch has learnt the routes via BGP for the remote hosts. The hosts that are local are also listed and are denoted with “i” for their path.

VXLAN EVPN With VRF

As we already have the VLANs configured, the VRF can just be added in.

Leaf 1
Leaf 2
Testing

As there is a VRF for these networks, it is easier to see the routes in a neater way.

The MAC address in the BGP route matches that of the local VXLAN interface

The MAC address in the BGP route matches that of the local VXLAN interface, the same for leaf 2 as leaf 1

VRF Route Leaking

Route leaking is allowing routes from one VRF to be added to another. This is useful for shared services. So all VRFs can have access to a default gateway or other services. In this example, I have simply added a new VRF to leaf 1 with two new VLANs (21 and 31). I have then permitted VLAN 20 into VRF CUSTOMER2 and VLAN 21 into VRF CUSTOMER1, allowing them to communicate, whilst being in different VRFs.

Leaf 1

Create the new VRF

Import the routes into the VRFs

Completed Configurations

Spine 1

Leaf 1

Leaf 2

Leave a Comment

Your email address will not be published. Required fields are marked *