Merge pull request #168 from klueska/create-cluster-with-cdi

Use CDI to inject GPUs into the kind workers instead of 'legacy' mode

README.md

@@ -2,7 +2,7 @@
 
 This DRA resource driver is currently under active development and not yet
 designed for production use.
-We will continually be force pushing over `main` until we have something more stable.
+We may (at times) decide to push commits over `main` until we have something more stable.
 Use at your own risk.
 
 A document and demo of the DRA support for GPUs provided by this repo can be found below:
@@ -32,7 +32,7 @@ First since we'll launch kind with GPU support, ensure that the following prereq
    Container Runtime to use volume mounts to select devices to inject into a
    container.
    ``` console
-   sudo nvidia-ctk config --set accept-nvidia-visible-devices-as-volume-mounts=true --in-place
+   sudo nvidia-ctk config --in-place --set accept-nvidia-visible-devices-as-volume-mounts=true
    ```
 
 1. Show the current set of GPUs on the machine:

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

@@ -28,7 +28,3 @@ kind create cluster \
 	--name "${KIND_CLUSTER_NAME}" \
 	--image "${KIND_IMAGE}" \
 	--config "${KIND_CLUSTER_CONFIG_PATH}"
-
-# Unmount the masked /proc/driver/nvidia to allow
-# dynamically generated MIG devices to be discovered
-docker exec -it "${KIND_CLUSTER_NAME}-worker" umount -R /proc/driver/nvidia

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

@@ -54,7 +54,7 @@ nodes:
   # 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/all
+    containerPath: /var/run/nvidia-container-devices/cdi/runtime.nvidia.com/gpu/all
   # The generated CDI specification assumes that `nvidia-ctk` is available on a
   # node -- specifically for the `nvidia-ctk hook` subcommand. As a workaround,
   # we mount it from the host.