Installation

This section describes the minimal settings required to install NVIDIA Base Command Manager on the primary head node. While the options selected here should be customized to suit the target environment during installation, many settings can also be reconfigured post install.

For more detailed information, please refer to the installation manual.

Details required to complete this section, with example values used in this guide:

Setting	Example Value
Cluster Name	nvidia-era-bcm-cluster
Organization Name	NVIDIA
Time zone	(GMT-07:00) America/Los_Angeles
Name servers	192.168.255.25, 192.168.255.26
Search domains	nvidia.com
Hostname	nvidia-era-bcm-01
Base IP address	192.168.100.0
Netmask	255.255.255.0(/24)
Dynamic range start	192.168.100.101
Dynamic range end	192.168.100.250
Gateway	192.168.100.1
Head node interface	ens224
IP address	192.168.100.253
IP offset	0.0.0.20
Installation drives	/dev/nvme0n1

Note

Where applicable these values will be displayed in red. Be sure to replace all example values with actual information relevant to the target environment.

1. Power on and boot the primary head node server from the installation ISO.

2. Select Start Base Command Manager Graphical Installer.

   

   Figure 2 GRUB boot menu with Start Base Command Manager Graphical Installer selected

3. Welcome Screen.

   a. Select Start Installation.

   

   Figure 3 Welcome screen with the Start Installation button

4. Software Licenses. NVIDIA EULA.

   a. Check the I agree checkbox.

   b. Click Next to continue.

   

   Figure 4 NVIDIA EULA software license agreement screen

5. Software License. Ubuntu Server EULA.

   a. Check the I agree checkbox.

   b. Click Next to continue.

   

   Figure 5 Ubuntu Server EULA software license agreement screen

6. Kernel Modules Configuration.

   a. Click Next to continue.

   

   Figure 6 Kernel Modules Configuration screen

   Note

   The Kernel Modules screen displays the modules recommended for loading based on a hardware probe. In most cases, the default selections can be accepted without adding additional modules.

7. Hardware Info.

   a. Click Next to continue.

   

   Figure 7 Hardware Info screen showing detected hardware

   Note

   The Hardware Info screen provides an overview of detected hardware based on the hardware probe used to load kernel modules. If any hardware is not detected at this stage, use the Back button to return to the Kernel Modules screen, add the appropriate modules, and then revisit the Hardware Info screen to verify detection.

8. Installation Source.

   a. Select the installation media from the list.

   b. Click Next to continue.

   

   Figure 8 Installation Source screen showing available installation media

9. Cluster Settings.

   a. Enter information for the target environment:

   i. Cluster name: nvidia-era-bcm-cluster

   ii. Organization name: NVIDIA

   iii. Time zone: (GMT-07:00) America/Los_Angeles

   iv. Name servers: 192.168.255.25 192.168.255.26

   v. Search domains: nvidia.com

   b. Click Next to continue.

   

   Figure 9 Cluster Settings screen with cluster name, organization, time zone, name servers, and search domains

10. Workload Management Configuration.

    a. Select None.

    b. Click Next to continue.

    

    Figure 10 Workload Management Configuration screen with None selected

    Note

    If no workload manager is selected here, one can be installed later after completing the cluster installation without it. Instructions for installing a workload manager post-installation are provided in the BCM Administrator Manual.

11. Network Topology.

    a. Select Type 2.

    b. Click Next to continue.

    

    Figure 11 Network Topology screen with Type 2 selected

    Note

    The design provided is based on the deployment of a type 2 network - a type 2 network has the compute nodes connected via a router to a public network.

    For more information on the different network topologies supported by BCM, please refer to the BCM Administrator Manual.

12. Head node Settings.

    a. Enter the following information:

    i. Hostname: nvidia-era-bcm-01

    ii. Password: <Password>

    b. Click Next to continue.

    

    Figure 12 Head node Settings screen with hostname and password fields

13. Compute Nodes Settings.

    a. Accept the default settings.

    b. Click Next to continue.

    

    Figure 13 Compute Nodes Settings screen with default settings

14. BMC Configuration.

    a. Accept the default settings.

    b. Click Next to continue.

    

    Figure 14 BMC Configuration screen with default settings

15. Networks.

    a. Ensure Internalnet is selected.

    b. Enter the following information:

    i. Base IP address: 192.168.100.0

    ii. Netmask: 255.255.255.0(/24)

    iii. Dynamic range start: 192.168.100.101

    iv. Dynamic range end: 192.168.100.250

    v. Gateway: 192.168.100.1

    c. Click Next to proceed.

    

    Figure 15 Networks screen showing Internalnet configuration with IP address range settings

    Note

    As per the NVIDIA Enterprise Reference Architecture, the first 20 IP addresses are reserved for network services.

    Setting the dynamic range start and end values suggested allows for other “static” or “reserved” IP addresses to be used in the range 192.168.100.21-192.168.100.100.

    The compute nodes utilize the DHCP assigned range 192.168.100.101-192.168.100.250.

    The BCM head nodes and High Availability shared IP address use the range 192.168.100.251-192.168.100-254.

16. Head node Interfaces.

    a. Select the appropriate interface.

    Note

    This should be set to the first interface of the DPU. In this example, the interface is listed as ens224, but in the target environment it should be enumerated in a pattern similar to enpXXs0f0np0.

    b. Set the IP address: 192.168.100.253

    Note

    It is suggested: use at least -1 from the last address in the usable range to allow for High Availability. For example: 192.168.100.253.

    c. Click Next to continue.

    

    Figure 16 Head node Interfaces screen showing interface selection and IP address configuration

17. Compute node interfaces.

    a. Set the IP offset: 0.0.0.20

    b. Click Next to continue.

    

    Figure 17 Compute node interfaces screen with IP offset set to 0.0.0.20

    Note

    Setting the IP offset to 0.0.0.20 configures the IP address of the first compute node to 192.168.100.21. Later in this guide, this will be changed to use a DHCP assigned IP address.

18. Disk Layout.

    a. Select the appropriate installation drives: /dev/nvme0n1

    Note

    In this example, there is only one drive listed, but in the target environment there may be multiple. It is recommended to select the hardware RAID configured disk.

    b. Click Next to proceed.

    

    Figure 18 Disk Layout screen showing available installation drives

19. Disk Layout Settings.

    a. Head node disk layout:

    i. Select One big partition. This setting cannot be not easily changed after installation.

    b. Compute node disk layout:

    i. Select One big partition. This will be further configured after installation.

    c. Click Next to continue.

    

    Figure 19 Disk Layout Settings screen with One big partition selected for both head node and compute node

20. Summary.

    a. Review all configuration settings to ensure they are correct.

    b. Click Start to begin the installation.

    

    Figure 20 Summary screen showing all configuration settings before installation begins

21. Deployment.

    a. Check the Automatically reboot after installation is complete checkbox.

    

    Figure 21 Deployment screen showing installation progress with automatic reboot option