Nvidia

Prerequisites

• Flux-CD
• Execute these commands to install the NVIDIA drivers and container toolkit on the host machine. This will allow the GPU Operator to manage the NVIDIA GPUs on the cluster.
• Add the NVIDIA package repositories

  curl -fsSL https://nvidia.github.io/libnvidia-container/gpgkey | sudo gpg --dearmor -o /usr/share/keyrings/nvidia-container-toolkit-keyring.gpg \
    && curl -s -L https://nvidia.github.io/libnvidia-container/stable/deb/nvidia-container-toolkit.list | \
      sed 's#deb https://#deb [signed-by=/usr/share/keyrings/nvidia-container-toolkit-keyring.gpg] https://#g' | \
      sudo tee /etc/apt/sources.list.d/nvidia-container-toolkit.list

  sudo sed -i -e '/experimental/ s/^#//g' /etc/apt/sources.list.d/nvidia-container-toolkit.list

  • Update the package lists

  sudo apt-get update

  • Install packages

  sudo apt install nvidia-driver nvidia-container-toolkit nvidia-container-toolkit-base

  • Configure the container runtime to use the NVIDIA runtime.

  sudo nvidia-ctk runtime configure --runtime=containerd

Installation

1. Create the following directory structure for NVIDIA GPU Operator:

   nvidia-gpu-operator/
   ├── gpu-operator.yml
   ├── gpu-operator-policy.yml
   └── gpu-operator-repo.yml

2. Add the following content to nvidia-gpu-operator/gpu-operator.yml:

   ---
   apiVersion: helm.toolkit.fluxcd.io/v2
   kind: HelmRelease
   metadata:
     name: gpu-operator
     namespace: gpu-operator
   spec:
     interval: 6h
     chart:
       spec:
         chart: gpu-operator
         version: "v25.10.1"
         sourceRef:
           kind: HelmRepository
           name: nvidia
           namespace: flux-system
         interval: 6h
     install:
       createNamespace: true
     upgrade:
       remediation:
         remediateLastFailure: true
     values:
       driver:
         enabled: false
       toolkit:
         env:
         - name: CONTAINERD_SOCKET
           value: /run/k3s/containerd/containerd.sock
         - name: CONTAINERD_CONFIG
           value: /var/lib/rancher/k3s/agent/etc/containerd/config.toml

3. Add the following content to nvidia-gpu-operator/gpu-operator-policy.yml:

   ---
   apiVersion: nvidia.com/v1
   kind: ClusterPolicy
   metadata:
     annotations:
       meta.helm.sh/release-name: gpu-operator
       meta.helm.sh/release-namespace: gpu-operator
     generation: 2
     labels:
       app.kubernetes.io/component: gpu-operator
       app.kubernetes.io/instance: gpu-operator
       app.kubernetes.io/managed-by: Helm
       app.kubernetes.io/name: gpu-operator
       app.kubernetes.io/version: v25.3.2
       helm.sh/chart: gpu-operator-v25.3.2
       helm.toolkit.fluxcd.io/name: gpu-operator
       helm.toolkit.fluxcd.io/namespace: gpu-operator
     name: cluster-policy
   spec:
     ccManager:
       defaultMode: "off"
       enabled: false
       env: []
       image: k8s-cc-manager
       imagePullPolicy: IfNotPresent
       repository: nvcr.io/nvidia/cloud-native
       version: v0.1.1
     cdi:
       default: false
       enabled: false
     daemonsets:
       labels:
         app.kubernetes.io/managed-by: gpu-operator
         helm.sh/chart: gpu-operator-v25.3.2
       priorityClassName: system-node-critical
       rollingUpdate:
         maxUnavailable: "1"
       tolerations:
       - effect: NoSchedule
         key: nvidia.com/gpu
         operator: Exists
       updateStrategy: RollingUpdate
     dcgm:
       enabled: false
       image: dcgm
       imagePullPolicy: IfNotPresent
       repository: nvcr.io/nvidia/cloud-native
       version: 4.2.3-1-ubuntu22.04
     dcgmExporter:
       enabled: true
       env:
       - name: DCGM_EXPORTER_LISTEN
         value: :9400
       - name: DCGM_EXPORTER_KUBERNETES
         value: "true"
       - name: DCGM_EXPORTER_COLLECTORS
         value: /etc/dcgm-exporter/dcp-metrics-included.csv
       image: dcgm-exporter
       imagePullPolicy: IfNotPresent
       repository: nvcr.io/nvidia/k8s
       serviceMonitor:
         additionalLabels: {}
         enabled: false
         honorLabels: false
         interval: 15s
         relabelings: []
       version: 4.2.3-4.1.3-ubuntu22.04
     devicePlugin:
       config:
         default: any
         name: time-slicing-config
       enabled: true
       env:
       - name: PASS_DEVICE_SPECS
         value: "true"
       - name: FAIL_ON_INIT_ERROR
         value: "true"
       - name: DEVICE_LIST_STRATEGY
         value: envvar
       - name: DEVICE_ID_STRATEGY
         value: uuid
       - name: NVIDIA_VISIBLE_DEVICES
         value: all
       - name: NVIDIA_DRIVER_CAPABILITIES
         value: all
       image: k8s-device-plugin
       imagePullPolicy: IfNotPresent
       repository: nvcr.io/nvidia
       version: v0.17.3
     driver:
       certConfig:
         name: ""
       enabled: false
       image: driver
       imagePullPolicy: IfNotPresent
       kernelModuleConfig:
         name: ""
       licensingConfig:
         configMapName: ""
         nlsEnabled: true
       manager:
         env:
         - name: ENABLE_GPU_POD_EVICTION
           value: "true"
         - name: ENABLE_AUTO_DRAIN
           value: "false"
         - name: DRAIN_USE_FORCE
           value: "false"
         - name: DRAIN_POD_SELECTOR_LABEL
           value: ""
         - name: DRAIN_TIMEOUT_SECONDS
           value: 0s
         - name: DRAIN_DELETE_EMPTYDIR_DATA
           value: "false"
         image: k8s-driver-manager
         imagePullPolicy: IfNotPresent
         repository: nvcr.io/nvidia/cloud-native
         version: v0.8.0
       rdma:
         enabled: false
         useHostMofed: false
       repoConfig:
         configMapName: ""
       repository: nvcr.io/nvidia
       startupProbe:
         failureThreshold: 120
         initialDelaySeconds: 60
         periodSeconds: 10
         timeoutSeconds: 60
       upgradePolicy:
         autoUpgrade: true
         drain:
           deleteEmptyDir: false
           enable: false
           force: false
           timeoutSeconds: 300
         maxParallelUpgrades: 1
         maxUnavailable: 25%
         podDeletion:
           deleteEmptyDir: false
           force: false
           timeoutSeconds: 300
         waitForCompletion:
           timeoutSeconds: 0
       useNvidiaDriverCRD: false
       usePrecompiled: false
       version: 570.148.08
       virtualTopology:
         config: ""
     gdrcopy:
       enabled: false
       image: gdrdrv
       imagePullPolicy: IfNotPresent
       repository: nvcr.io/nvidia/cloud-native
       version: v2.5
     gfd:
       enabled: true
       env:
       - name: GFD_SLEEP_INTERVAL
         value: 60s
       - name: GFD_FAIL_ON_INIT_ERROR
         value: "true"
       image: k8s-device-plugin
       imagePullPolicy: IfNotPresent
       repository: nvcr.io/nvidia
       version: v0.17.3
     hostPaths:
       driverInstallDir: /run/nvidia/driver
       rootFS: /
     kataManager:
       config:
         artifactsDir: /opt/nvidia-gpu-operator/artifacts/runtimeclasses
         runtimeClasses:
         - artifacts:
             pullSecret: ""
             url: nvcr.io/nvidia/cloud-native/kata-gpu-artifacts:ubuntu22.04-535.54.03
           name: kata-nvidia-gpu
           nodeSelector: {}
         - artifacts:
             pullSecret: ""
             url: nvcr.io/nvidia/cloud-native/kata-gpu-artifacts:ubuntu22.04-535.86.10-snp
           name: kata-nvidia-gpu-snp
           nodeSelector:
             nvidia.com/cc.capable: "true"
       enabled: false
       image: k8s-kata-manager
       imagePullPolicy: IfNotPresent
       repository: nvcr.io/nvidia/cloud-native
       version: v0.2.3
     mig:
       strategy: single
     migManager:
       config:
         default: all-disabled
         name: default-mig-parted-config
       enabled: true
       env:
       - name: WITH_REBOOT
         value: "false"
       gpuClientsConfig:
         name: ""
       image: k8s-mig-manager
       imagePullPolicy: IfNotPresent
       repository: nvcr.io/nvidia/cloud-native
       version: v0.12.2-ubuntu20.04
     nodeStatusExporter:
       enabled: false
       image: gpu-operator-validator
       imagePullPolicy: IfNotPresent
       repository: nvcr.io/nvidia/cloud-native
       version: v25.3.2
     operator:
       defaultRuntime: docker
       initContainer:
         image: cuda
         imagePullPolicy: IfNotPresent
         repository: nvcr.io/nvidia
         version: 12.8.1-base-ubi9
       runtimeClass: nvidia
     psa:
       enabled: false
     sandboxDevicePlugin:
       enabled: true
       image: kubevirt-gpu-device-plugin
       imagePullPolicy: IfNotPresent
       repository: nvcr.io/nvidia
       version: v1.3.1
     sandboxWorkloads:
       defaultWorkload: container
       enabled: false
     toolkit:
       enabled: true
       env:
       - name: CONTAINERD_SOCKET
         value: /run/k3s/containerd/containerd.sock
       - name: CONTAINERD_CONFIG
         value: /var/lib/rancher/k3s/agent/etc/containerd/config.toml
       image: container-toolkit
       imagePullPolicy: IfNotPresent
       installDir: /usr/local/nvidia
       repository: nvcr.io/nvidia/k8s
       version: v1.17.8-ubuntu20.04
     validator:
       image: gpu-operator-validator
       imagePullPolicy: IfNotPresent
       plugin:
         env:
         - name: WITH_WORKLOAD
           value: "false"
       repository: nvcr.io/nvidia/cloud-native
       version: v25.3.2
     vfioManager:
       driverManager:
         env:
         - name: ENABLE_GPU_POD_EVICTION
           value: "false"
         - name: ENABLE_AUTO_DRAIN
           value: "false"
         image: k8s-driver-manager
         imagePullPolicy: IfNotPresent
         repository: nvcr.io/nvidia/cloud-native
         version: v0.8.0
       enabled: true
       image: cuda
       imagePullPolicy: IfNotPresent
       repository: nvcr.io/nvidia
       version: 12.8.1-base-ubi9
     vgpuDeviceManager:
       config:
         default: default
         name: ""
       enabled: true
       image: vgpu-device-manager
       imagePullPolicy: IfNotPresent
       repository: nvcr.io/nvidia/cloud-native
       version: v0.3.0
     vgpuManager:
       driverManager:
         env:
         - name: ENABLE_GPU_POD_EVICTION
           value: "false"
         - name: ENABLE_AUTO_DRAIN
           value: "false"
         image: k8s-driver-manager
         imagePullPolicy: IfNotPresent
         repository: nvcr.io/nvidia/cloud-native
         version: v0.8.0
       enabled: false
       image: vgpu-manager
       imagePullPolicy: IfNotPresent

4. Add the following content to nvidia-gpu-operator/gpu-operator-repo.yml:

   ---
   apiVersion: source.toolkit.fluxcd.io/v1
   kind: HelmRepository
   metadata:
     name: nvidia
     namespace: flux-system
   spec:
     interval: 6h
     url: https://helm.ngc.nvidia.com/nvidia

5. Commit and push these files to your Git repository. The Flux CD will automatically detect the changes and deploy the NVIDIA GPU Operator to your cluster.