Merge pull request #54826 from mindprince/addon-manager

Automatic merge from submit-queue (batch tested with PRs 54826, 53576, 55591, 54946, 54825). If you want to cherry-pick this change to another branch, please follow the instructions <a href="https://github.com/kubernetes/community/blob/master/contributors/devel/cherry-picks.md">here</a>.

Run nvidia-gpu device-plugin daemonset as an addon on GCE nodes that have nvidia GPUs attached

- Instead of the old `Accelerators` feature that added `alpha.kubernetes.io/nvidia-gpu` resource, use the new `DevicePlugins` feature that adds vendor specific resources. (In case of nvidia GPUs it will
add `nvidia.com/gpu` resource.)

- Add node label to GCE nodes with accelerators attached. This node label is the same as what GKE attaches to node pools with accelerators attached. (For example, for nvidia-tesla-p100 GPU, the label would be `cloud.google.com/gke-accelerator=nvidia-tesla-p100`) This will help us target accelerator specific
daemonsets etc. to these nodes.

- Run nvidia-gpu device-plugin daemonset as an addon on GCE nodes that have nvidia GPUs attached.

- Some minor documentation improvements in addon manager.

**Release note**:
```release-note
GCE nodes with NVIDIA GPUs attached now expose `nvidia.com/gpu` as a resource instead of `alpha.kubernetes.io/nvidia-gpu`.
```

/sig cluster-lifecycle
/sig scheduling
/area hw-accelerators

https://github.com/kubernetes/features/issues/368

cluster/common.sh

@@ -877,6 +877,11 @@ EOF
     if [ -n "${CLUSTER_SIGNING_DURATION:-}" ]; then
       cat >>$file <<EOF
 CLUSTER_SIGNING_DURATION: $(yaml-quote ${CLUSTER_SIGNING_DURATION})
+EOF
+    fi
+    if [[ "${NODE_ACCELERATORS:-}" == *"type=nvidia"* ]]; then
+      cat >>$file <<EOF
+ENABLE_NVIDIA_GPU_DEVICE_PLUGIN: $(yaml-quote "true")
 EOF
     fi