Space DRA Driver (Proof of Concept)

This repository contains a proof of concept for a DRA (Dynamic Resource Allocation) driver that provisions namespaces on the same cluster as the driver. It's forked from the dra-example-driver project.

The intention is to demonstrate how any resource, not just devices/hardware, can be dynamically provisioned with the necessary artifacts made available securely to running containers (through env variables, mounts, etc.)

Quickstart and Demo

The driver itself provides access to namespaces created as requested, and this demo walks through the process of building and installing the driver followed by running a set of workloads that request these namespaces.

The procedure below has been tested and verified on Linux (Fedora).

Prerequisites

• GNU Make 3.81+
• GNU Tar 1.34+
• docker v20.10+ (including buildx)
• kind v0.17.0+
• helm v3.7.0+
• kubectl v1.18+

Demo

This demo requires a specific directory be present on the host:

mkdir /tmp/claim-artifacts

Create a kind cluster to run everything in:

./demo/create-cluster.sh

Build the image for the example resource driver:

./demo/build-driver.sh

And then install the example resource driver via helm:

helm upgrade -i \
  --create-namespace \
  --namespace dra-example-driver \
  dra-example-driver \
  deployments/helm/dra-example-driver

Double check the driver components have come up successfully:

$ kubectl get pod -n dra-example-driver
NAME                                             READY   STATUS    RESTARTS   AGE
dra-example-driver-controller-7555d488db-nbd52   1/1     Running   0          1m
dra-example-driver-kubeletplugin-qwmbl           1/1     Running   0          1m

Next, deploy an example app that demonstrates how two pods can share a dynamically provisioned namespace:

kubectl apply --filename=demo/namespace-test.yaml

And verify that they come up successfully:

$ kubectl get pods -n namespace-test
NAME   READY   STATUS    RESTARTS   AGE
pod0   1/1     Running   0          1m
pod1   1/1     Running   0          1m

Now take a look at the namespaces on the cluster. Take note of the namespace prefixed with ephemeral-ns-.

$ kubectl get namespaces
NAME                 STATUS   AGE
default              Active   10m
dra-example-driver   Active   8m
ephemeral-ns-4rsv8   Active   1m
kube-node-lease      Active   10m
kube-public          Active   10m
kube-system          Active   10m
local-path-storage   Active   10m
namespace-test       Active   1m

Print the environment variables of the containers to see what was injected by the driver:

$ kubectl exec -n namespace-test pod0 -- printenv
DRA_RESOURCE_DRIVER_NAME=space.resource.example.com
TEST_CLAIM_CLUSTER=https://cluster.todo
TEST_CLAIM_NAMESPACE=ephemeral-ns-4rsv8
TEST_CLAIM_KUBECONFIG=/var/run/claim-artifacts/e2140c9f-4fea-44a6-b688-2f52763194ab/kubeconfig
...

$ kubectl exec -n namespace-test pod1 -- printenv
DRA_RESOURCE_DRIVER_NAME=space.resource.example.com
TEST_CLAIM_CLUSTER=https://cluster.todo
TEST_CLAIM_NAMESPACE=ephemeral-ns-4rsv8
TEST_CLAIM_KUBECONFIG=/var/run/claim-artifacts/e2140c9f-4fea-44a6-b688-2f52763194ab/kubeconfig
...

Likewise, print the contents of the kubeconfig (not yet implemented) mounted to each container:

$ kubectl exec -n namespace-test pod0 -- cat /etc/test-claim/kubeconfig        
TODO%

$ kubectl exec -n namespace-test pod1 -- cat /etc/test-claim/kubeconfig        
TODO%

To see namespace cleanup in action, delete the example app:

kubectl delete --filename=demo/namespace-test.yaml

Notice how the ephemeral namespace is gone:

$ kubectl get namespaces
NAME                 STATUS   AGE
default              Active   12m
dra-example-driver   Active   10m
kube-node-lease      Active   12m
kube-public          Active   12m
kube-system          Active   12m
local-path-storage   Active   12m

Take a look at the driver logs for more insight into the process:

kubectl logs -n dra-example-driver -l app.kubernetes.io/name=dra-example-driver

Delete your cluster when you're done.

./demo/delete-cluster.sh

References

For more information on the DRA Kubernetes feature and developing custom resource drivers, see the following resources:

• Dynamic Resource Allocation in Kubernetes
• KEP: 3063-dynamic-resource-allocation