Driver persistence changes

api/nvidia/v1alpha1/nvidiadriver_types.go

@@ -59,6 +59,12 @@ type NVIDIADriverSpec struct {
 	// +operator-sdk:gen-csv:customresourcedefinitions.specDescriptors.x-descriptors="urn:alm:descriptor:com.tectonic.ui:booleanSwitch"
 	UseOpenKernelModules *bool `json:"useOpenKernelModules,omitempty"`
 
+	// PersistDriver indicates if the driver install should be persisted across restarts
+	PersistDriver *bool `json:"persist,omitempty"`
+
+	// InstallDirectory is the install location for the driver
+	InstallDirectory string `json:"installDirectory,omitempty"`
+
 	// NVIDIA Driver container startup probe settings
 	StartupProbe *ContainerProbeSpec `json:"startupProbe,omitempty"`
 

config/crd/bases/nvidia.com_nvidiadrivers.yaml

@@ -204,6 +204,9 @@ spec:
                 items:
                   type: string
                 type: array
+              installDirectory:
+                description: InstallDirectory is the install location for the driver
+                type: string
               kernelModuleConfig:
                 description: 'Optional: Kernel module configuration parameters for
                   the NVIDIA Driver'
@@ -511,6 +514,10 @@ spec:
                 description: NodeSelector specifies a selector for installation of
                   NVIDIA driver
                 type: object
+              persist:
+                description: PersistDriver indicates if the driver install should
+                  be persisted across restarts
+                type: boolean
               priorityClassName:
                 description: 'Optional: Set priorityClassName'
                 type: string

config/samples/nvidia_v1alpha1_nvidiadriver.yaml

@@ -18,6 +18,8 @@ spec:
     useHostMofed: false
   gds:
     enabled: false
+  persist: false
+  installDirectory: '/opt/nvidia/driver'
   # Private mirror repository configuration
   repoConfig:
     name: ""

deployments/gpu-operator/crds/nvidia.com_nvidiadrivers.yaml

@@ -204,6 +204,9 @@ spec:
                 items:
                   type: string
                 type: array
+              installDirectory:
+                description: InstallDirectory is the install location for the driver
+                type: string
               kernelModuleConfig:
                 description: 'Optional: Kernel module configuration parameters for
                   the NVIDIA Driver'
@@ -511,6 +514,10 @@ spec:
                 description: NodeSelector specifies a selector for installation of
                   NVIDIA driver
                 type: object
+              persist:
+                description: PersistDriver indicates if the driver install should
+                  be persisted across restarts
+                type: boolean
               priorityClassName:
                 description: 'Optional: Set priorityClassName'
                 type: string

manifest.diff

@@ -0,0 +1,51 @@
+diff --git a/manifests/state-driver/0500_daemonset.yaml b/manifests/state-driver/0500_daemonset.yaml
+index 8ceb7820c..8716d147d 100644
+--- a/manifests/state-driver/0500_daemonset.yaml
++++ b/manifests/state-driver/0500_daemonset.yaml
+@@ -205,6 +205,12 @@ spec:
+         # always use runc for driver containers
+         - name: NVIDIA_VISIBLE_DEVICES
+           value: void
++      {{- if .Driver.Spec.PersistDriver }}
++        - name: PERSIST_DRIVER
++          value: "true"
++        - name: INSTALL_DIR
++          value: {{ .Driver.Spec.InstallDirectory | default "/opt/nvidia/driver" }}
++      {{- end }}
+       {{- if deref .Driver.Spec.UseOpenKernelModules }}
+         - name: OPEN_KERNEL_MODULES_ENABLED
+           value: "true"
+@@ -254,6 +260,14 @@ spec:
+         {{- end }}
+       {{- end }}
+         volumeMounts:
++          {{- if .Driver.Spec.PersistDriver }}
++          - name: install-dir
++            mountPath: {{ .Driver.Spec.InstallDirectory | default "/opt/nvidia/driver" }}
++          - name: lib-modules
++            mountPath: /lib/modules
++          - name: dev
++            mountPath: /dev
++          {{- end }}
+           - name: run-nvidia
+             mountPath: /run/nvidia
+             mountPropagation: Bidirectional
+@@ -574,6 +588,18 @@ spec:
+             readOnly: true
+       {{- end }}
+       volumes:
++        {{- if .Driver.Spec.PersistDriver }}
++        - name: install-dir
++          hostPath:
++            path: {{ .Driver.Spec.InstallDirectory | default "/opt/nvidia/driver" }}
++            type: DirectoryOrCreate
++        - name: lib-modules
++          hostPath:
++            path: /lib/modules
++        - name: dev
++          hostPath:
++            path: /dev
++        {{- end }}
+         - name: run-nvidia
+           hostPath:
+             path: /run/nvidia

manifests/state-driver/0500_daemonset.yaml

@@ -205,6 +205,12 @@ spec:
         # always use runc for driver containers
         - name: NVIDIA_VISIBLE_DEVICES
           value: void
+      {{- if .Driver.Spec.PersistDriver }}
+        - name: PERSIST_DRIVER
+          value: "true"
+        - name: INSTALL_DIR
+          value: {{ .Driver.Spec.InstallDirectory | default "/opt/nvidia/driver" }}
+      {{- end }}
       {{- if deref .Driver.Spec.UseOpenKernelModules }}
         - name: OPEN_KERNEL_MODULES_ENABLED
           value: "true"
@@ -254,6 +260,14 @@ spec:
         {{- end }}
       {{- end }}
         volumeMounts:
+          {{- if .Driver.Spec.PersistDriver }}
+          - name: install-dir
+            mountPath: {{ .Driver.Spec.InstallDirectory | default "/opt/nvidia/driver" }}
+          - name: lib-modules
+            mountPath: /lib/modules
+          - name: dev
+            mountPath: /dev
+          {{- end }}
           - name: run-nvidia
             mountPath: /run/nvidia
             mountPropagation: Bidirectional
@@ -574,6 +588,18 @@ spec:
             readOnly: true
       {{- end }}
       volumes:
+        {{- if .Driver.Spec.PersistDriver }}
+        - name: install-dir
+          hostPath:
+            path: {{ .Driver.Spec.InstallDirectory | default "/opt/nvidia/driver" }}
+            type: DirectoryOrCreate
+        - name: lib-modules
+          hostPath:
+            path: /lib/modules
+        - name: dev
+          hostPath:
+            path: /dev
+        {{- end }}
         - name: run-nvidia
           hostPath:
             path: /run/nvidia