Add new nvkind target for demo/clusters with multinode GPUs support

The new nvkind target assumes you have nvkind installed on your system. If you
do, it will plit all of the GPUs onto individual nodes and spin up a multi-node
kind cluster with 1 GPU per node.

Signed-off-by: Kevin Klues <[email protected]>

demo/clusters/kind/install-dra-driver.sh

@@ -22,8 +22,8 @@ set -o pipefail
 
 source "${CURRENT_DIR}/scripts/common.sh"
 
-kubectl label node "${KIND_CLUSTER_NAME}-worker" --overwrite nvidia.com/dra.kubelet-plugin=true
-kubectl label node "${KIND_CLUSTER_NAME}-control-plane" --overwrite nvidia.com/dra.controller=true
+kubectl label node -l node-role.x-k8s.io/worker --overwrite nvidia.com/dra.kubelet-plugin=true
+kubectl label node -l node-role.x-k8s.io/control-plane --overwrite nvidia.com/dra.controller=true
 
 helm upgrade -i --create-namespace --namespace nvidia-dra-driver nvidia ${PROJECT_DIR}/deployments/helm/k8s-dra-driver \
     ${NVIDIA_DRIVER_ROOT:+--set nvidiaDriverRoot=${NVIDIA_DRIVER_ROOT}} \

demo/clusters/kind/scripts/kind-cluster-config.yaml

@@ -25,6 +25,8 @@ containerdConfigPatches:
     enable_cdi = true
 nodes:
 - role: control-plane
+  labels:
+    node-role.x-k8s.io/control-plane: ""
   kubeadmConfigPatches:
   - |
     kind: ClusterConfiguration
@@ -43,6 +45,8 @@ nodes:
       kubeletExtraArgs:
         v: "1"
 - role: worker
+  labels:
+    node-role.x-k8s.io/worker: ""
   kubeadmConfigPatches:
   - |
     kind: JoinConfiguration

demo/clusters/nvkind/build-dra-driver.sh

@@ -0,0 +1 @@
+../kind/build-dra-driver.sh

demo/clusters/nvkind/create-cluster.sh

@@ -0,0 +1 @@
+../kind/create-cluster.sh

demo/clusters/nvkind/delete-cluster.sh

@@ -0,0 +1 @@
+../kind/delete-cluster.sh

demo/clusters/nvkind/install-dra-driver.sh

@@ -0,0 +1 @@
+../kind/install-dra-driver.sh

demo/clusters/nvkind/scripts/build-driver-image.sh

@@ -0,0 +1 @@
+../../kind/scripts/build-driver-image.sh

demo/clusters/nvkind/scripts/build-kind-image.sh

@@ -0,0 +1 @@
+../../kind/scripts/build-kind-image.sh

demo/clusters/nvkind/scripts/common.sh

@@ -0,0 +1 @@
+../../kind/scripts/common.sh

demo/clusters/nvkind/scripts/create-kind-cluster.sh

@@ -0,0 +1,30 @@
+#!/usr/bin/env bash
+
+# Copyright 2023 The Kubernetes Authors.
+# Copyright 2023 NVIDIA CORPORATION.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+# A reference to the current directory where this script is located
+CURRENT_DIR="$(cd -- "$( dirname -- "${BASH_SOURCE[0]}" )" &> /dev/null && pwd)"
+
+set -ex
+set -o pipefail
+
+source "${CURRENT_DIR}/common.sh"
+
+nvkind cluster create \
+	--retain \
+	--name "${KIND_CLUSTER_NAME}" \
+	--image "${KIND_IMAGE}" \
+	--config-template "${KIND_CLUSTER_CONFIG_PATH}"

demo/clusters/nvkind/scripts/delete-kind-cluster.sh

@@ -0,0 +1 @@
+../../kind/scripts/delete-kind-cluster.sh

demo/clusters/nvkind/scripts/kind-cluster-config.yaml

@@ -0,0 +1,63 @@
+# Copyright 2023 The Kubernetes Authors.
+# Copyright 2023 NVIDIA CORPORATION.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+kind: Cluster
+apiVersion: kind.x-k8s.io/v1alpha4
+featureGates:
+  DynamicResourceAllocation: true
+containerdConfigPatches:
+# Enable CDI as described in
+# https://tags.cncf.io/container-device-interface#containerd-configuration
+- |-
+  [plugins."io.containerd.grpc.v1.cri"]
+    enable_cdi = true
+nodes:
+- role: control-plane
+  labels:
+    node-role.x-k8s.io/control-plane: ""
+  kubeadmConfigPatches:
+  - |
+    kind: ClusterConfiguration
+    apiServer:
+        extraArgs:
+          runtime-config: "resource.k8s.io/v1alpha3=true"
+    scheduler:
+        extraArgs:
+          v: "1"
+    controllerManager:
+        extraArgs:
+          v: "1"
+  - |
+    kind: InitConfiguration
+    nodeRegistration:
+      kubeletExtraArgs:
+        v: "1"
+{{- range $gpu := until numGPUs }}
+- role: worker
+  labels:
+    node-role.x-k8s.io/worker: ""
+  kubeadmConfigPatches:
+  - |
+    kind: JoinConfiguration
+    nodeRegistration:
+      kubeletExtraArgs:
+        v: "1"
+  extraMounts:
+  # We inject all NVIDIA GPUs using the nvidia-container-runtime.
+  # This requires `accept-nvidia-visible-devices-as-volume-mounts = true` be set
+  # in `/etc/nvidia-container-runtime/config.toml`
+  - hostPath: /dev/null
+    containerPath: /var/run/nvidia-container-devices/cdi/runtime.nvidia.com/gpu/{{ $gpu }}
+{{- end }}

demo/clusters/nvkind/scripts/load-driver-image-into-kind.sh

@@ -0,0 +1 @@
+../../kind/scripts/load-driver-image-into-kind.sh