Configure NSX host transport node on DPU-based vSphere Lifecycle Manager-enabled cluster

You need to configure NSX and enable Enhanced Datapath for vSphere Distributed Services Engine to offload some of the network operations from your server CPU to the DPU.

Configuring NSX host transport node on a DPU-based vSphere Lifecycle Manager-enabled cluster is similar to preparing the existing NSX cluster on vSphere Lifecycle Manager.

vSphere Distributed Switch (VDS) backed by the DPU on ESXi supports the offloading mode after NSX is enabled. Traffic forwarding logic is offloaded from the ESXi host to the VDS backed by DPU.

Note: A DPU is supported only on vSphere Lifecycle Managed clusters. DPU requires a minimum version combination of vSphere 8.0, NSX 4.0.1.1, and Edge hardware version 20.

To learn more about Network Offloads Capability, see What is Network Offloads Capability in the VMware vSphere Documentation.

Prerequisites

• vSphere offloads: DPUs are supported on hosts in a vSphere Lifecycle Manager-enabled cluster that are running on ESXi 8.0 version or higher.
  Note: For vSphere offload capabilities, you do not need to purchase a separate NSX license. You just need to have the vSphere ENT + Term license and be on vSphere 8. NSX Manager is available as a part of vSphere ENT+.
• NSX offloads: To utilize the NSX DPU-based acceleration capabilities, you need to purchase an NSX Enterprise Plus Term license (Per Core basis) or an NSX Enterprise Plus with Threat Prevention Term license (Per Core basis).
  Note: You do not need any additional NSX licenses if you have an NSX Enterprise Plus Term license (Per Core) or an NSX Enterprise Plus with Threat Prevention Term license (Per Core).
  Note: vSphere ENT+ license holders on vSphere 8.0 are entitled to 'Enhanced Datapath - Standard for DPUs'. This does not require you to purchase a separate NSX license.
  For more information, see NSX Feature and Edition Guide.
• Lockdown mode is not enabled on any of the hosts. vSphere Lifecycle Manager might fail to prepare hosts that are enabled to function in Lockdown mode.
• There is not drift in images between hosts and cluster. Otherwise, you cannot enable vSphere Lifecycle Manager on the cluster. Remediate hosts in vCenter Server to ensure base image matches on host and cluster.
• vSphere Lifecycle Manager is enabled on the cluster. See VMware vSphere^® documentation.
• Register Compute Manager with the following settings:
  • Enable Trust and set access level to vSphere Lifecycle Manager. Trust is mandatory to establish communication between NSX and vSphere Lifecycle Manager.
  • Enable Create Service Account.
• Create a transport node profile using a vSphere Distributed Switch host switch. NSX Virtual Distributed Switch (N-VDS) is not supported on vSphere Lifecycle Managed clusters.

Procedure

1. From a browser, log in with admin privileges to an NSX Manager at https://<nsx-manager-ip-address> or https://<nsx-manager-fqdn>.
2. Go to System → Fabric → Nodes → Host Transport Nodes.
3. In the Managed by drop-down menu, select the Compute Manager.
4. Select the cluster and click Configure NSX.
   Note: Identify vSphere Lifecycle Manager-enabled cluster when a cluster is accompanied with vLCM text.
5. To create a transport node profile, click Add Profile.
6. Enter Transport Node Profile (TNP) details as required.
   Note:

   • For VDS backed by the DPU, select the Host TN and TNP Host Switch mode as Enhanced Datapath - Standard (Recommended) or Enhanced Datapath - Performance.
   • In Add Switch, select VDS (Distributed Switch version 8.0 created in vSphere Client) with offload compatibility. If there is a mismatch, the host will not be compatible.
   • One cluster allows only one type of host. For example, when you create two distributed switches on vSphere client: one with NVIDIA network offload compatibility and one with AMD Pensando network offload compatibility, both options will appear on the UI. Select the one specific to your requirement. Clusters of different types cannot use the same TNP. Therefore, you will need two separate TNPs: one for NVIDIA DPU and one for AMD Pensando DPU.
7. If TNP is already created, select a transport node profile that uses vSphere Distributed Switch as the host switch and datapath mode as Enhanced Datapath.
8. Click Apply TNP.

   If this is the first cluster that is enabled for vSphere Lifecycle Manager, NSX uploads the NSX LCP bundle to the image repository in vCenter Server. vSphere Lifecycle Manager sets NSX as a solution on the cluster. It sets the desired state to the NSX image uploaded to vCenter Server. Then, vSphere Lifecycle Manager begins installation of NSX VIBs on each host, followed by configuration of NSX switch on each transport node.

   As part of host preparation, vSphere Lifecycle Manager remediates the host, registers the host with NSX Manager, configures NSX switch on the host and completes the configuration.

   Note:

   • vSphere Lifecycle Manager puts the ESXi host backed by AMD Pensando DPU in maintenance mode and reboots it as part of host remediation. If vSphere Lifecycle Manager fails to place the host in maintenance mode, you need to manually power off all VMs and then retry NSX installation
   • Installing NSX on a vSphere Lifecycle Manager-enabled cluster might take a little more time than when installing on a non-vSphere Lifecycle Manager-enabled cluster. This difference is due to the additional health checks that are included in this combination of products.