Merge pull request #459 from sky1122/enable-cdi-k8s-containerd

Support both CDI and Legacy NVIDIA Container Runtime modes

Author: arnaldo2792

packages/nvidia-container-toolkit/nvidia-container-toolkit-config-k8s

@@ -1,14 +1,39 @@
 [required-extensions]
 nvidia-container-runtime = "v1"
-std = { version = "v1", helpers = ["default"] }
+kubelet-device-plugins = "v1"
+std = { version = "v1", helpers = ["default", "is_array"] }
 
 +++
 ### generated from the template file ###
 accept-nvidia-visible-devices-as-volume-mounts = {{default true settings.nvidia-container-runtime.visible-devices-as-volume-mounts}}
 accept-nvidia-visible-devices-envvar-when-unprivileged = {{default false settings.nvidia-container-runtime.visible-devices-envvar-when-unprivileged}}
-
 [nvidia-container-cli]
 root = "/"
 path = "/usr/bin/nvidia-container-cli"
 environment = []
 ldconfig = "@/sbin/ldconfig"
+
+[nvidia-container-runtime]
+{{#if settings.kubelet-device-plugins.nvidia.device-list-strategy}}
+{{~#if (is_array settings.kubelet-device-plugins.nvidia.device-list-strategy) ~}}
+{{~#if (eq settings.kubelet-device-plugins.nvidia.device-list-strategy.[0] "cdi-cri") ~}}
+mode="cdi"
+{{~else~}}
+mode="legacy"
+{{~/if~}}
+{{~else~}}
+{{~#if (eq settings.kubelet-device-plugins.nvidia.device-list-strategy "cdi-cri") ~}}
+mode="cdi"
+{{~else~}}
+mode="legacy"
+{{~/if~}}
+{{/if}}
+{{else}}
+mode="legacy"
+{{/if}}
+
+[nvidia-container-runtime-hook]
+# For the legacy NVIDIA runtime, skip detecting the mode used in the
+# NVIDA Container Runtime. This prevents failures in the legacy NVIDIA runtime
+# when the selected mode is 'cdi'.
+skip-mode-detection = true

packages/nvidia-k8s-device-plugin/nvidia-k8s-device-plugin-conf

@@ -1,6 +1,6 @@
 [required-extensions]
 kubelet-device-plugins = "v1"
-std = { version = "v1", helpers = ["default"] }
+std = { version = "v1", helpers = ["default", "join_array", "is_array"] }
 
 +++
 version: v1
@@ -15,10 +15,20 @@ flags:
   migStrategy: "none"
   {{/if}}
   failOnInitError: true
+  nvidiaDriverRoot: "/"
   plugin:
     passDeviceSpecs: {{default true settings.kubelet-device-plugins.nvidia.pass-device-specs}}
-    deviceListStrategy: {{default "volume-mounts" settings.kubelet-device-plugins.nvidia.device-list-strategy}}
+    {{#if settings.kubelet-device-plugins.nvidia.device-list-strategy}}
+    {{#if (is_array settings.kubelet-device-plugins.nvidia.device-list-strategy)}}
+    deviceListStrategy: [{{join_array "," settings.kubelet-device-plugins.nvidia.device-list-strategy }}]
+    {{else}}
+    deviceListStrategy: {{settings.kubelet-device-plugins.nvidia.device-list-strategy}}
+    {{/if}}
+    {{else}}
+    deviceListStrategy: "volume-mounts"
+    {{/if}}
     deviceIDStrategy: {{default "index" settings.kubelet-device-plugins.nvidia.device-id-strategy}}
+    containerDriverRoot: "/"
 {{#if settings.kubelet-device-plugins.nvidia.device-sharing-strategy}}
 {{#if (eq settings.kubelet-device-plugins.nvidia.device-sharing-strategy "time-slicing")}}
 sharing: