Enterprise Manageability#
Enterprise IT teams that operate DGX Spark systems at scale can use the following sections for fleet lifecycle integration and for custom installation when removable media or on-premises mirrors replace direct use of public package sources.
Enterprise Lifecycle Integration
- Enterprise Lifecycle Integration
- How to Use This Information
- Support Boundaries and Operational Responsibilities
- DGX Spark Overview, Enterprise Challenges, Lifecycle Backbone, and the JSON/Agentless SSH
- Integration Patterns: SSH Execution from Management Platforms
- Implementation
- Clear Asset Information (Seven Tools)
- Controlled Software/Firmware Updates (One Tool)
- Remote Ops and Remediation (Two Tools)
- Tool Locations and Reference Code Map
- Platform Wrapper Patterns and Packaging Examples
Custom Installation with Cloud-Init
- Custom Installation with Cloud-Init
- Overview
- Air-Gapped and Custom Installation Patterns
- Example Constants
- Customize the BaseOS Image with repack_baseos.sh
- USB Partitioning and the OEMDATA Layout
- Host a Minimal APT Repository and Firmware Tree
- Mirror the Full Ubuntu Ports and LVFS Content on a Server
- Client Configuration and hook.sh
- Security Considerations
- Verify the Customization and Installation Outcomes
- Prepare the Installation Media and Client (Verification Flow)
- Reference: OEM Scripts and Cloud-Init
- Validation Scenarios and Feedback Questions
By accessing the instructions for custom installation and scaled deployment, the Customer acknowledges that this workflow bypasses individual end-user license prompts. Use of these tools is governed by the NVIDIA Software License Agreement (which includes the NVIDIA CUDA EULA). By proceeding, the Administrator represents they have the authority to bind the Customer to these terms and conditions.