You can run artificial intelligence (AI) and machine learning (ML) workloads on vSphere Bitfusion 2.5 without losing your current cluster configuration and monitoring data.

To upgrade your cluster, you must back up the environment, deploy new server virtual machines (VMs) with version 2.5 of the vSphere Bitfusion appliance, and restore the backup.

You can use your current vSphere Bitfusion 2.0 clients or upgrade the clients to version 2.5. To upgrade a client, you must install the latest CentOS, Red Hat, or Ubuntu package. For more information, see Installing the vSphere Bitfusion Client.
Figure 1. vSphere Bitfusion Upgrade Workflow
Workflow diagram of the vSphere Bitfusion upgrade process.
  1. By using the vSphere Bitfusion Plug-in, create and download a backup of your vSphere Bitfusion 2.0 cluster.

    For more information, see Back up a Bitfusion Cluster in the VMware vSphere Bitfusion User Guide.

  2. From the downloaded bitfusionbackup.tar.gz archive file, open the manifest.json file, and find the servers section. The section includes information about the servers in your vSphere Bitfusion cluster at the time of the backup.

    Take note of the host ID, hostname, and number of GPUs for each server.

    For example, the host ID of the following server is 6a2f4e80-70d8-4c51-bf10-00284f3ed2c6, the hostname is bitfusion-server-2.0.1-3-1, and the server has one GPU installed.
    "servers": [
            {
                "id": "6a2f4e80-70d8-4c51-bf10-00284f3ed2c6",
                "hostname": "bitfusion-server-2.0.1-3-1",
                "ip": "10.202.8.209",
                "port": "56001",
                "address": "10.202.8.209:56001",
                "mode": "manager",
                "health": "PASS",
                "num_devices": 1,
                "lastseen": "2020-10-14T21:29:38Z",
                "license": {
                    "type": "vcenter-license",
                    "name": "vSphere 7 Enterprise Plus",
                    "license-id": "example",
                    "key": "example",
                    "expiry": "2025-09-30T00:00:00Z"
                }
  3. Power off all vSphere Bitfusion servers in the cluster.
  4. Install a new primary vSphere Bitfusion server.
    1. Deploy a new primary vSphere Bitfusion server VM by using a vSphere Bitfusion 2.5 Appliance OVF Template.

      For more information, see Deploying the vSphere Bitfusion Appliance.

      During the deployment process, enter the same hostname as your primary vSphere Bitfusion 2.0 server uses.

    2. In the settings of the new VM, add the same number of GPUs as your primary vSphere Bitfusion 2.0 server uses.
    3. In the settings of the new VM, add a guestinfo.bitfusion.server.host-id configuration parameter. The parameter value must match the host ID of your primary server with version 2.0, that is listed in the manifest.json file.

      For more information, see Edit Configuration File Parameters in the vSphere Virtual Machine Administration documentation.

    4. Power on the server and wait until the vSphere Bitfusion Plug-in is registered with vCenter Server.
  5. By using the vSphere Bitfusion Plug-in, restore the backup of your vSphere Bitfusion 2.0 cluster.

    For more information, see Restore a Bitfusion Cluster in the VMware vSphere Bitfusion User Guide.

  6. For each subsequent vSphere Bitfusion server in your cluster, perform the following steps.
    1. Deploy a new server VM by using a vSphere Bitfusion 2.5 Appliance OVF Template.

      During the deployment process, enter the same hostname as the corresponding vSphere Bitfusion 2.0 server uses.

    2. In the settings of the new VM, add the same number of GPUs as the corresponding vSphere Bitfusion 2.0 server uses.
    3. In the settings of the new VM, add a guestinfo.bitfusion.server.host-id configuration parameter. The parameter value must match the host ID of the corresponding server with version 2.0, that is listed in the manifest.json file.
    4. Enable the VM as a vSphere Bitfusion server.

      For more information, see Add Subsequent vSphere Bitfusion Servers.

    5. Power on the VM. Multiple VMs must be powered on in a sequential order.
  7. Delete the vSphere Bitfusion 2.0 server VMs.

The servers in your cluster are upgraded to version 2.5.