NVIDIA DGX Explained: SuperPOD / BasePOD Update Guide

Update Guide

• Overview
• Scope
• Update and Upgrade Best Practices
• SuperPOD / BasePOD Software Minor Updates Using BCM
  • High-Level Steps
  • BCM Headnode Updates
• DGX OS
  • Update DGX OS
• Update Individual Software Packages With BCM
  • CUDA Toolkit
  • DCGM
  • Enroot
  • GPU Driver
  • MOFED (Mellanox OFED) Managed by BCM – (Preferred Method)
  • MOFED (Mellanox OFED) Not Managed by BCM
  • DOCA OFED Transition
• Slurm Updates
• Kubernetes Updates
  • Prerequisites
  • Control Plane Node Updates
  • Worker Node Updates
  • Operator Updates
  • Run:ai Updates
• DGX Firmware Packages
  • Importing NVFWUPD
• InfiniBand Components
  • UFM Appliance Update
  • InfiniBand Switch Upgrade
  • Determine Upgrade Path
  • Initiate the Upgrade
  • InfiniBand Transceivers/Cable Firmware Upgrade
• NVIDIA Spectrum Ethernet Switches
  • Follow the Following Steps if the Switch Is NOT Using MLAG
  • Follow the Following Steps if the Switch Is Using MLAG
• Reference Links

Notices

• Notices
  • Notice
  • Trademarks
  • Copyright