Install Using Base Command Manager (BCM)

NVIDIA Base Command Manager (BCM) provides an automated installation path for NVIDIA Run:ai on DGX SuperPOD and BasePOD systems. From BCM 11.33.1, the BCM cm-kubernetes-setup wizard provisions Kubernetes and configures the required components (GPU Operator, Network Operator, MetalLB, Kgateway, Prometheus), after which the dedicated cm-runai-setup wizard deploys the NVIDIA Run:ai control plane and cluster.

This section covers everything needed to complete a BCM-based NVIDIA Run:ai deployment:

• Preparations - Infrastructure requirements that must be in place before deployment: IP address reservations, DNS records, TLS certificates, registry credentials, and BCM node categories. Applies to both connected and air-gapped deployments.
• Connected deployment - Step-by-step walkthrough of both installation wizards for internet-connected clusters.
• Air-Gapped deployment - End-to-end deployment guide for clusters with no internet access.

The DGX platform configuration pages (B200, B300, GB200, GB300) contain hardware-specific files required during deployment. Validation tests provide post-deployment checks to confirm the installation is functioning correctly.