Merge branch 'cherry-picks-for-v0.12.3' into 'release-0.12'

Cherry-picks from master into release-0.12 branch for v0.12.3

See merge request nvidia/kubernetes/device-plugin!207

Author: klueska

README.md

@@ -111,7 +111,7 @@ version = 2
           runtime_root = ""
           runtime_type = "io.containerd.runc.v2"
           [plugins."io.containerd.grpc.v1.cri".containerd.runtimes.nvidia.options]
-            BinaryName = "/usr/bin/nvidia-container-runtime
+            BinaryName = "/usr/bin/nvidia-container-runtime"
 ```
 And then restart `containerd`:
 ```

deployments/container/Dockerfile.ubi8

@@ -27,6 +27,9 @@ RUN make PREFIX=/artifacts cmds
 
 FROM nvidia/${CUDA_IMAGE}:${CUDA_VERSION}-base-${BASE_DIST}
 
+# Remove CUDA libs(compat etc) in favor of libs installed by the NVIDIA driver
+RUN dnf remove -y cuda-*
+
 ENV NVIDIA_DISABLE_REQUIRE="true"
 ENV NVIDIA_VISIBLE_DEVICES=all
 ENV NVIDIA_DRIVER_CAPABILITIES=utility

deployments/container/Dockerfile.ubuntu

@@ -27,6 +27,9 @@ RUN make PREFIX=/artifacts cmds
 
 FROM nvidia/${CUDA_IMAGE}:${CUDA_VERSION}-base-${BASE_DIST}
 
+# Remove CUDA libs(compat etc) in favor of libs installed by the NVIDIA driver
+RUN apt-get --purge -y autoremove cuda-*
+
 ENV NVIDIA_DISABLE_REQUIRE="true"
 ENV NVIDIA_VISIBLE_DEVICES=all
 ENV NVIDIA_DRIVER_CAPABILITIES=utility

deployments/helm/nvidia-device-plugin/charts/gpu-feature-discovery/templates/_helpers.tpl

@@ -0,0 +1,10 @@
+{{/*
+Allow the release namespace to be overridden for multi-namespace deployments in combined charts
+*/}}
+{{- define "gpu-feature-discovery.namespace" -}}
+  {{- if .Values.namespaceOverride -}}
+    {{- .Values.namespaceOverride -}}
+  {{- else -}}
+    {{- .Release.Namespace -}}
+  {{- end -}}
+{{- end -}}

deployments/helm/nvidia-device-plugin/charts/gpu-feature-discovery/templates/validation.yml

@@ -0,0 +1,8 @@
+{{- if and (eq (include "gpu-feature-discovery.namespace" .) "default") ( eq .Values.namespaceOverride "") (not .Values.allowDefaultNamespace) }}
+{{- $error := "" }}
+{{- $error = printf "%s\nRunning in the 'default' namespace is not recommended." $error }}
+{{- $error = printf "%s\nSet 'allowDefaultNamespace=true' to bypass this error." $error }}
+{{- $error = printf "%s\nOtherwise, use --namespace (with --create-namespace as necessary) to run in a specific namespace." $error }}
+{{- $error = printf "%s\nSee: https://helm.sh/docs/helm/helm_install/#options" $error }}
+{{- fail $error }}
+{{- end }}

deployments/helm/nvidia-device-plugin/charts/gpu-feature-discovery/values.yaml

@@ -3,6 +3,7 @@ sleepInterval: null
 
 nameOverride: ""
 fullnameOverride: ""
+namespaceOverride: ""
 selectorLabelsOverride: {}
 
 imagePullSecrets: []

deployments/helm/nvidia-device-plugin/templates/_helpers.tpl

@@ -24,6 +24,17 @@ If release name contains chart name it will be used as a full name.
 {{- end -}}
 {{- end -}}
 
+{{/*
+Allow the release namespace to be overridden for multi-namespace deployments in combined charts
+*/}}
+{{- define "nvidia-device-plugin.namespace" -}}
+  {{- if .Values.namespaceOverride -}}
+    {{- .Values.namespaceOverride -}}
+  {{- else -}}
+    {{- .Release.Namespace -}}
+  {{- end -}}
+{{- end -}}
+
 {{/*
 Create chart name and version as used by the chart label.
 */}}
@@ -79,17 +90,17 @@ Security context for the plugin
 */}}
 {{- define "nvidia-device-plugin.securityContext" -}}
 {{- if ne (len .Values.securityContext) 0 -}}
-  {{ toYaml .Values.securityContext }}
+{{ toYaml .Values.securityContext }}
 {{- else if .Values.compatWithCPUManager -}}
-  privileged: true
+privileged: true
 {{- else if ne (include "nvidia-device-plugin.allPossibleMigStrategiesAreNone" .) "true" -}}
-    capabilities:
-      add:
-        - SYS_ADMIN
+capabilities:
+  add:
+    - SYS_ADMIN
 {{- else -}}
-  allowPrivilegeEscalation: false
-  capabilities:
-    drop: ["ALL"]
+allowPrivilegeEscalation: false
+capabilities:
+  drop: ["ALL"]
 {{- end -}}
 {{- end -}}
 
@@ -98,15 +109,15 @@ Security context for GFD
 */}}
 {{- define "gpu-feature-discovery.securityContext" -}}
 {{- if ne (len .Subcharts.gfd.Values.securityContext) 0 -}}
-  {{ toYaml .Subcharts.gfd.Values.securityContext }}
+{{ toYaml .Subcharts.gfd.Values.securityContext }}
 {{- else if ne (include "nvidia-device-plugin.allPossibleMigStrategiesAreNone" .) "true" -}}
-    capabilities:
-      add:
-        - SYS_ADMIN
+capabilities:
+  add:
+    - SYS_ADMIN
 {{- else -}}
-  allowPrivilegeEscalation: false
-  capabilities:
-    drop: ["ALL"]
+allowPrivilegeEscalation: false
+capabilities:
+  drop: ["ALL"]
 {{- end -}}
 {{- end -}}
 

deployments/helm/nvidia-device-plugin/templates/configmap.yml

@@ -3,6 +3,7 @@ apiVersion: v1
 kind: ConfigMap
 metadata:
   name: {{ include "nvidia-device-plugin.configMapName" . }}
+  namespace: {{ include "nvidia-device-plugin.namespace" . }}
   labels:
     {{- include "nvidia-device-plugin.labels" . | nindent 4 }}
 data:

deployments/helm/nvidia-device-plugin/templates/daemonset.yml

@@ -24,6 +24,7 @@ apiVersion: apps/v1
 kind: DaemonSet
 metadata:
   name: {{ include "nvidia-device-plugin.fullname" . }}
+  namespace: {{ include "nvidia-device-plugin.namespace" . }}
   labels:
     {{- include "nvidia-device-plugin.labels" . | nindent 4 }}
 spec:

deployments/helm/nvidia-device-plugin/templates/gfd.yml

@@ -24,6 +24,7 @@ apiVersion: apps/v1
 kind: DaemonSet
 metadata:
   name: {{ include "nvidia-device-plugin.fullname" . }}
+  namespace: {{ include "nvidia-device-plugin.namespace" . }}
   labels:
     {{- include "nvidia-device-plugin.labels" . | nindent 4 }}
 spec: