GPU Operator on OpenShift
- Introduction
- Prerequisites
- Overview
- Installing the Node Feature Discovery (NFD) Operator
- Installing the NVIDIA GPU Operator
- Installing the NVIDIA GPU Operator by using the web console
- Installing the NVIDIA GPU Operator using the CLI
- Create the ClusterPolicy instance
- Create the ClusterPolicy instance with NVIDIA vGPU
- Verify the successful installation of the NVIDIA GPU Operator
- Cluster monitoring
- Logging
- Running a sample GPU Application
- Getting information about the GPU
- NVIDIA AI Enterprise with OpenShift
- MIG Support in OpenShift Container Platform
- Cleanup
- Deploy GPU Operators in a disconnected or airgapped environment
- Introduction
- Prerequisites
- Set up a basic HTTP Server
- Optional: Check the version of RHEL being used in the cluster
- Optional: Mirror the RPM packages
- Creating a private registry
- Authenticate the mirror registry
- Configuring credentials that allow images to be mirrored
- Mirror the Operator catalogs on a disconnected cluster
- Disabling the default OperatorHub sources
- Pruning an index image
- Mirror Node Feature Discovery and the NVIDIA GPU Operator Catalog
- Creating a catalog from an index image
- Verify the mirrored catalog source
- Install the Node Feature Discovery Operator
- Optional: Installing the NVIDIA GPU Operator on OpenShift version
4.8.19,4.8.21,4.9.8 - Install the NVIDIA GPU Operator
- Enable the GPU Operator Dashboard
- Time-slicing NVIDIA GPUs in OpenShift
- NVIDIA GPU Operator with OpenShift Virtualization
- Introduction
- Assumptions, constraints, and dependencies
- Prerequisites
- Labeling worker nodes
- Building the vGPU Manager image
- Installing the NVIDIA GPU Operator using the CLI
- Create the secret
- Creating a ClusterPolicy for the GPU Operator
- Enabling the IOMMU driver on hosts
- Add GPU resources to the HyperConverged CR
- Creating a virtual machine with GPU
- vGPU Device Configuration
- Troubleshooting
- Appendix