You can add additional servers to your vSphere Bitfusion cluster when you require more GPU resources.

After the primary vSphere Bitfusion server starts, vSphere Bitfusion registers a vSphere Bitfusion plug-in in the vCenter Server, resulting in a single vSphere Bitfusion cluster containing one vSphere Bitfusion server. After the vSphere Bitfusion plug-in is registered, you can add subsequent servers by following the steps in this procedure. The vSphere Bitfusion plug-in uses the primary server's configuration data, which allows faster deployment of the subsequent servers.

Alternatively, you can add a new server in your vSphere Bitfusion cluster by following the deploy procedure for the primary server. You deploy the vSphere Bitfusion appliance on a virtual machine (VM), customize the vSphere Bitfusion OVF template, pass through the GPUs to the vSphere Bitfusion server VM, and enable the VM as a vSphere Bitfusion server.

Additional vSphere Bitfusion servers must be part of the same vCenter Server instance as the first vSphere Bitfusion server.

Prerequisites

  • Verify you have installed a primary vSphere Bitfusion server.
  • Verify that the vSphere Bitfusion is registered with vCenter Server server.

Procedure

  1. From the Hosts and Clusters view in vCenter Server, right-click an ESXi host, and select Bitfusion > Install Bitfusion server.
    The Install Bitfusion server dialog box appears.
  2. On the Select an OVA image page, enter the URL of the vSphere Bitfusion OVA file or browse to the file, and click Next.
  3. On the Verify template details page, review the OVA template details and click Next.
  4. On the Select a name and hostname page, enter a name for the virtual machine and a hostname for the vSphere Bitfusion server, and click Next.
    Optionally, you can specify a host ID for the vSphere Bitfusion server, for example, when you upgrade your vSphere Bitfusion server. If you skip this step, a host ID is generated and assigned automatically.
  5. On the Select storage page, define where and how to store the files of the deployed VM, and click Next.
  6. On the Select networks page, specify the networking configuration for Network Adapter 1 and click Next.
    You must specify the configuration for Network Adapter 1 that is used for management and data traffic. Network Adapter 1 must be connected to a network that communicates with the vCenter Server instance.
    If your vSphere Bitfusion server requires additional network adapters for data traffic, you can click Add Network Adapter and specify the network configuration for the additional adapter.
    Option Description
    Network Adapter Select a network from the drop-down menu.
    Adapter Type Select a network adapter to assign to the virtual machine.
    Note: vSphere Bitfusion supports VMXNET3 and PVRDMA adapters.
    DHCP/Fixed IP Specify whether a DHCP server assigns the address of the network adapter or you use a fixed IPv4 address.
    IPv4 Address Enter the IPv4 address of the network adapter. If you are using DHCP, leave this text box blank.
    Note: IPv6 is not supported.
    Netmask Select a netmask from the drop-down menu.

    For example, if your network uses a /24 netmask, choose 24 (255.255.255.0).

    .
    Gateway Enter the network gateway address to use with the appliance. If you are using DHCP, leave this text box blank.
    MTU Enter an MTU size. The default value is 1500. For optimal performance, specify an MTU size of 4000 or greater.
    Note: If you set an MTU size greater than 1500, verify that the network switches in your data center are enabled for jumbo frames.
    DNS Servers Enter the DNS server address to use with the appliance. If you are using DHCP, leave this text box blank.
    DNS Search Domains Enter the DNS search domain address to use with the appliance. If you are using DHCP, leave this text box blank.
    NTP Enter the NTP server address to use with the appliance. If you are using DHCP and the DHCP server supports sending NTP server information, leave this text box blank.
  7. On the Select GPUs page, add GPUs to the subsequent server and click Next.
    1. Click Add GPU.
    2. Select a GPU from the GPU Device drop-down menu.
    3. (Optional) Specify the total memory of the GPU.
    4. (Optional) To accept the NVIDIA license, select the Download and Install NVIDIA Driver check box.
      By accepting the NVIDIA license, vSphere Bitfusion downloads and installs the NVIDIA driver, CUDA libraries, and NVIDIA Fabric Manager during the first boot of the virtual machine.
      Note: If you are operating vSphere Bitfusion in an environment without access to the Internet, for example, by using an air-gapped network, do not select the check box. You must manually download and install the NVIDIA software after deploying the vSphere Bitfusion appliance.
    If your vSphere Bitfusion server requires additional GPUs, you can click Add GPU Device again and specify the settings for the GPU.
  8. On the Customize server page, specify the vSphere Bitfusion server details and click Next.
    1. Specify the number of CPUs for the virtual machine.
    2. Specify the total virtual machine memory in GB.
    3. (Optional) Enter a password for the customer account.
      After the deployment is complete, you use the customer user account to log into the vSphere Bitfusion server by using the console shell or SSH. If you skip this step, you cannot log into the subsequent server.
    4. (Optional) Select the Power On VM After Create check box.
      You can deselect the check box, if you make changes to the virtual machine before powering it on.
  9. On the Summary page, review the deployment details and click Finish.

Results

A new task for installing the vSphere Bitfusion server appears in the Recent Tasks pane. After the task finishes, the new appliance is created on the selected resources.

When a new vSphere Bitfusion server joins the cluster, vCenter Server supplies a token, a certificate, and a configuration to access the vSphere Bitfusion cluster.