Garvit/aws gpu test

.gitignore

@@ -27,3 +27,8 @@ config/**/charts
 *.swp
 *.swo
 *~
+
+# Terraform files
+*.tfstate
+*.tfstate.backup
+.terraform*

Makefile

@@ -125,11 +125,11 @@ help: ## Display this help.
 
 .PHONY: manifests
 manifests: controller-gen ## Generate WebhookConfiguration, ClusterRole and CustomResourceDefinition objects.
-	$(CONTROLLER_GEN) rbac:roleName=manager-role crd webhook paths="./..." output:crd:artifacts:config=config/crd/bases
+	$(CONTROLLER_GEN) rbac:roleName=manager-role crd webhook paths="./..." output:crd:artifacts:config=config/crd/bases -w
 
 .PHONY: generate
 generate: controller-gen ## Generate code containing DeepCopy, DeepCopyInto, and DeepCopyObject method implementations.
-	$(CONTROLLER_GEN) object:headerFile="hack/boilerplate.go.txt" paths="./..."
+	$(CONTROLLER_GEN) object:headerFile="hack/boilerplate.go.txt" paths="./..." -w
 
 .PHONY: fmt
 fmt: ## Run go fmt against code.

dist/install.yaml

@@ -1229,4 +1229,4 @@ spec:
       volumes:
       - configMap:
           name: devzero-zxporter-env-config
-        name: config-volume
+        name: config-volume

nvidia-device-plugin-prereq/container-toolkit.yaml

@@ -0,0 +1,84 @@
+apiVersion: apps/v1
+kind: DaemonSet
+metadata:
+  name: nvidia-toolkit-installer
+  namespace: nvidia-device-plugin
+spec:
+  selector:
+    matchLabels:
+      name: nvidia-toolkit-installer
+  template:
+    metadata:
+      labels:
+        name: nvidia-toolkit-installer
+    spec:
+      nodeSelector:
+        nvidia.com/gpu.present: "true"
+      hostPID: true
+      tolerations:
+        - key: "nvidia.com/gpu"
+          operator: "Exists"
+          effect: "NoSchedule"
+        - key: "CriticalAddonsOnly"
+          operator: "Exists"
+        - effect: NoSchedule
+          key: node-role.kubernetes.io/control-plane
+        - effect: NoSchedule
+          key: node-role.kubernetes.io/master
+      containers:
+      - name: install-nvidia-toolkit
+        image: amazonlinux:2023
+        securityContext:
+          privileged: true
+        command:
+          - /bin/bash
+          - -c
+          - |
+            set -ex
+
+            # Add NVIDIA repo
+            curl -s -L https://nvidia.github.io/libnvidia-container/stable/rpm/nvidia-container-toolkit.repo \
+              -o /etc/yum.repos.d/nvidia-container-toolkit.repo
+
+            # Install toolkit
+            yum install -y nvidia-container-toolkit
+
+            # Configure containerd
+            nvidia-ctk runtime configure --runtime=containerd
+
+            # Restart containerd
+            systemctl restart containerd || true
+
+            # Exit cleanly
+            echo "NVIDIA container toolkit installed and configured."
+            sleep infinity
+        volumeMounts:
+        - name: root
+          mountPath: /host
+          mountPropagation: Bidirectional
+        - name: containerd-config
+          mountPath: /etc/containerd
+        - name: systemd
+          mountPath: /run/systemd
+        - name: modules
+          mountPath: /lib/modules
+          readOnly: true
+        - name: dev
+          mountPath: /dev
+      volumes:
+        - name: root
+          hostPath:
+            path: /
+        - name: containerd-config
+          hostPath:
+            path: /etc/containerd
+        - name: systemd
+          hostPath:
+            path: /run/systemd
+        - name: modules
+          hostPath:
+            path: /lib/modules
+        - name: dev
+          hostPath:
+            path: /dev
+      restartPolicy: Always

nvidia-device-plugin-prereq/driver-installer.yaml

@@ -0,0 +1,81 @@
+apiVersion: apps/v1
+kind: DaemonSet
+metadata:
+  name: nvidia-driver-installer
+  namespace: nvidia-device-plugin
+spec:
+  selector:
+    matchLabels:
+      name: nvidia-driver-installer
+  template:
+    metadata:
+      labels:
+        name: nvidia-driver-installer
+    spec:
+      hostPID: true
+      tolerations:
+        - key: nvidia.com/gpu
+          operator: Exists
+          effect: NoSchedule
+        - key: CriticalAddonsOnly
+          operator: Exists
+        - key: node-role.kubernetes.io/control-plane
+          effect: NoSchedule
+        - key: node-role.kubernetes.io/master
+          effect: NoSchedule
+      nodeSelector:
+        nvidia.com/gpu.present: "true"
+      containers:
+      - name: driver-installer
+        image: nvcr.io/nvidia/cloud-native/k8s-driver-manager:v0.8.0
+        securityContext:
+          privileged: true
+        env:
+          - name: NVIDIA_DRIVER_VERSION
+            value: "535.129.03"  # or the version you require
+          - name: NODE_NAME
+            valueFrom:
+              fieldRef:
+                fieldPath: spec.nodeName
+        volumeMounts:
+          - name: root
+            mountPath: /host
+            mountPropagation: Bidirectional
+          - name: modules
+            mountPath: /lib/modules
+            readOnly: true
+          - name: nvidia-local
+            mountPath: /host/usr/local/nvidia
+      - name: fix-dcgm-dir
+        image: amazonlinux:2023
+        securityContext:
+          privileged: true
+        command: ["/bin/bash", "-c"]
+        args:
+          - |
+            set -ex
+            TARGET_DIR="/host/usr/local/nvidia"
+            # If it doesn't exist, symlink something useful
+            if [ ! -d "$TARGET_DIR" ]; then
+              mkdir -p /host/usr/local
+              ln -s /usr/lib64 "$TARGET_DIR"
+            fi
+            echo "/usr/local/nvidia set up for DCGM."
+            sleep 10
+        volumeMounts:
+          - name: nvidia-local
+            mountPath: /host/usr/local/nvidia
+          - name: root
+            mountPath: /host
+            mountPropagation: Bidirectional
+      volumes:
+        - name: root
+          hostPath:
+            path: /
+        - name: modules
+          hostPath:
+            path: /lib/modules
+        - name: nvidia-local
+          hostPath:
+            path: /usr/local/nvidia
+            type: DirectoryOrCreate