Deploy NSX-T Edge Nodes
July 14, 2020NSX-T Edge nodes are used for security and gateway services that can’t be run on the distributed routers in use by NSX-T. These edge nodes do things like North/South routing, load balancing, DHCP, VPN, NAT, etc. If you want to use Tier0
or Tier1
routers, you will need to have at least 1 edge node deployed. These edge nodes provide a place to run services like the Tier0 routes. When you first deploy an edge, its like an empty shell of a VM until these services are needed.
In my lab, I’m deploying the edge nodes to their own cluster. This is not a requirement for a lab, but a good recommendation for a production setup since traffic is usually funneled through these instances and they can become a network hot spot.
Before we do the deploy, let’s revisit this logical diagram for the edge node we’ll be deploying. I’ll be honest, this edge networking caused me fits until I realized that we need to extend the overlay networks from the ESXi hosts in our workload cluster, to the Edge Virtual Machine. You must add the Edge VM to the TEP network to participate in the overlay. Then, you will have a second VM interface that connects to a VLAN transport zone which will be the portgroup created on my Edge ESXi host virtual switch.
To wrap your head around the Edge VM networking, take a look at this page, and specifically the diagram found below. The Edge VM has a virtual switch inside it, and we’ll connect the edge vm uplinks to the Distributed virtual switch uplinks. The Edge VM will have three or more interfaces. 1 for Management, 1 for Overlay, and 1 for VLAN traffic to the physical network.
In the end, our overlay network will extend from the ESXi hosts to the Edge virtual machine. The edge virtual machine will have a path to the physical network.
To deploy the first edge node, go to the NSX Manager under System –> Fabric –> Nodes -> Edge Transport Nodes. Click the + ADD EDGE VM
button.
Give the edge vm a name, FQDN, and description before selecting a size. Sizing is critically important for a production environment. It has impacts on the number of load balancers that can be provisioned among other things.
NOTE: The documentation states that you need the Large
form factor for the Tanzu components. I was able to get it to come up with a Medium
form factor, but could not deploy TKG clusters until I upgraded to Large
.
Next, select the compute manager, cluster, resource pool, and datastore for the edge node to be deployed on.
In the configure node settings box, give the VM an IP address on a management network. This is NOT in the data plane, but rather a way to communicate with the edge node. I’ve placed this VM on one of my existing management portgroups on my edge cluster. (VLAN 50 – Management)
The last screen is where the configuration really happens. This is where we create virtual switches in the Edge VM and connect them (through uplinks) to a physical Nic. We need to create two switches in this screen, these switches will not be visible in the ESXi hosts view, because they exist within the Edge virtual machine.
We need two switches created. One for the Overlay network which belongs on the TEP network along with our ESXi hosts in our workload cluster. And another for the VLAN backed network which is how the VM communicates on the physical network.
On the Configure NSX
screen, click the +ADD SWITCH
link twice so we can setup each switch. The first I’ve named nsx-vlan to represent the northbound physical network. I selected the VLAN-Zone transport zone (VLAN 201 – Edge Network) and the single nic profile which is an out-of-the-box profile. Under the teaming policy, I’ve selected my Edge Uplink portgroup that was already created on my ESXi DVS. This link IS within the data path.
On the second switch in the configuration I’ve added the Overlay-Zone transport zone, again with the single nic profile. Under Address assignment, select Use IP Pool and select the TEP pool that we also used on our workload ESXi hosts to add them to the overlay. Then select the uplink for the NSX-TEP network (VLAN 200 – NSX TEP).
I know this piece was confusing to me, so if you’re stuck, take a look at the diagram below. The edge VM will be created with a pair of switches. The uplinks on those switches will be portgroups on the DVS. My lab layout is shown below for the edge VM.
After finishing the configuration, an edge VM should be listed under your Edge Transport Nodes.
Create Edge Cluster
Edge Nodes can (and in production environments should) be deployed in pairs. Groups of Edge nodes can then be pooled together into an Edge cluster. Thus far, we haven’t focused enough on High Availability because its a lab short of resources, but these routines should be modified slightly to provide this HA capability. This might include adding a second VLAN transport zone for a second physical switch etc. Edge clusters, although we won’t be using more than one in our example, are required.
Add your new Edge node to an edge cluster by navigating to System –> Fabric –> Nodes –> Edge Transport Zones. Click the + ADD
button to create a new cluster.
Give the cluster a name and description. Then make sure your edge VM has been selected and moved to the right column.
Summary
I found that understanding the edge node routing was the most difficult piece to setting up NSX in the lab. Remember that we’re extending the Overlay transport zone from the workload ESXi host cluster to the Edge node VM. The Edge VM then has a second VLAN transport zone where traffic can be routed to the physical network. Stay tuned for the next post were we create some actual overlay networks that our VMs can use.
[…] and configured our profiles to configure the virtual switches and NICs on the ESXi nodes. In the next post we’ll move to our Edge […]
I think this is challenging for most of us the first time we discover it. Host TEP and Edge TEP on the same dvSwitch/vmnic is not supported, and VMware does not exactly go out of their way to point this out. Even worse, they will not admit to it being a bug in the kernel’s networking layer. Hopefully they will fix this in 3.1 or 3.5.
This is an excellent point. In my lab, I’m using the Host TEP and Edge TEP on the same VLAN. This is OK if they are in different clusters. IF you need to deploy an edge node on the same hosts running the overlay networks, you need to place the edges on their own portgroup and route to the host overlay network. This has bitten me in the past as well. Thank you for pointing it out in the comments.
OMG i have spent the last 2 nights trying to get my home lab hosts to connect to the edge. Totally missed the edge TEP needs to be on different DVS. been tearing down, re-building, banging head against desk
Thanks, glad I read this blog and the replies as was getting nowhere fast with the edges not working.
[…] 4 – Deploy NSX-T Edge Nodes […]
I created two new port groups on the same VDS: EdgeOverlayvlan27 and Edgeuplinkvlanall.On this basis, two new EDGE, EDGE3VS7 and EDGE3VS72 were created.
Name: EDGE3VS7
Host name/FQDN: EDGE3VS7.tt.com,
Large size:
IP allocation: static
IP: 192.168.25.46/24
GW: Such as 192.168.25.1,
Management interface: VMNETWORK
N-VDS1:switch-overlay
Transport area: TZ-overlay
Upstream profile: NSX-Edge-single-NIC-Uplink-Profile
IP allocation: Use IP Pool
: IP pool edge
Uplink-1 / Edgeoverlayvlan27
N-VDS2:switch-vlan
Transport area: TZ-VLAN
Upstream profile: NSX-Edge-single-NIC-Uplink-Profile
Uplink-1 / Edgeuplinkvlanall
Name: EDGE3VS72
Host name/FQDN: EDGE3VS72.tt.com,
Large size:
IP allocation: static
IP: 192.168.25.53/24
GW: Such as 192.168.25.1,
Management interface: VMNETWORK
N-VDS1:switch-overlay
Transport area: TZ-overlay
Upstream profile: NSX-Edge-single-NIC-Uplink-Profile
IP allocation: Use IP Pool
: IP pool edge
Uplink-1 / Edgeoverlayvlan27
N-VDS2:switch-vlan
Transport area: TZ-VLAN
Upstream profile: NSX-Edge-single-NIC-Uplink-Profile
Uplink-1 / Edgeuplinkvlanall
Now we encounter a problem: EDGE3VS7 has TWO VTEP channels with ESXI transmission node and EDGEVS72.However, there is only one VTEP channel with EDGE3VS7 on EDGEVS72, and no channel with ESXI transmission node.Why is that?
I am bet confused, when we deploy Edge it comes with 4 NICs one of them connected to the VLAN VDS-PG in the Vsphere “which is used for physical connectivity”, and in T0 GW we can add 2 interfaces attached to 2 segments type VLAN “”also is used for physical connectivity””. so could you please clarify the differed between them?