WIP: Add KEP-4680 device health reporting with message field support

[harche]

Summary

Adds KEP-4680 device health reporting to the example driver, serving as the reference implementation for the version-neutral WatchHealthStatus API from the kubeletplugin helper (kubernetes/kubernetes#139477). No versioned DRAResourceHealth gRPC types appear in driver code; the helper serves v1 and v1alpha1, so allocated device health surfaces in pod.status.containerStatuses[].allocatedResourcesStatus on both new and 1.36-era kubelets.

The driver reports all devices healthy and re-sends the report periodically as the heartbeat the kubelet's health lease requires. For demos and e2e tests, annotating the plugin pod with health.example.com/<device>: Unhealthy|Healthy|Unknown overrides a device's simulated health (see demo/examples/health-reporting/).

go.mod temporarily pins the k8s.io staging modules to master pseudo-versions until WatchHealthStatus ships in a k8s.io/dynamic-resource-allocation release.

Verification

The e2e test deploys a pod with a claim, asserts Healthy in pod status, flips the device to Unhealthy via annotation, and asserts recovery back to Healthy. Verified on kind against both supported kubelet generations:

• v1alpha1 path: released kindest/node:v1.36.1 (pre-v1 kubelet) — e2e passes via the helper's v1alpha1 compatibility serving.
• v1 path: node image built from kubernetes master (v1.37.0-beta.0.446+4b1379a4a44) — e2e passes, and driver logs confirm the kubelet opened /v1.DRAResourceHealth/NodeWatchResources.

[k8s-ci-robot]

Welcome @harche!

It looks like this is your first PR to kubernetes-sigs/dra-example-driver 🎉. Please refer to our pull request process documentation to help your PR have a smooth ride to approval.

You will be prompted by a bot to use commands during the review process. Do not be afraid to follow the prompts! It is okay to experiment. Here is the bot commands documentation.

You can also check if kubernetes-sigs/dra-example-driver has its own contribution guidelines.

You may want to refer to our testing guide if you run into trouble with your tests not passing.

If you are having difficulty getting your pull request seen, please follow the recommended escalation practices. Also, for tips and tricks in the contribution process you may want to read the Kubernetes contributor cheat sheet. We want to make sure your contribution gets all the attention it needs!

Thank you, and welcome to Kubernetes. 😃

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is NOT APPROVED

This pull-request has been approved by: harche
Once this PR has been reviewed and has the lgtm label, please assign klueska for approval. For more information see the Code Review Process.

The full list of commands accepted by this bot can be found here.

Details

Needs approval from an approver in each of these files:

• OWNERS

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[harche]

/hold for the implementation PRs to merge,

kubernetes/enhancements#4680 (comment)

[harche]

I will update once 1.36 beta is cut

[harche]

/cc @SergeyKanzhelev @nojnhuh

[harche]

I will update once 1.36 beta is cut

/hold cancel

[harche]

/cc @SergeyKanzhelev @nojnhuh

/hold I am adding a feature to trigger the health status update in this example driver.

[harche]

Putting this on hold until #201 merges. Both PRs modify the e2e test infrastructure (e2e_setup_test.go, e2e_test.go) and will conflict. Will rebase after #201 lands.

/hold

[pohly]

As mentioned before, I don't like that DRA drivers have to interact with the gRPC interface (drahealthv1alpha1 currently, despite the beta graduation of the feature?!). Can we first improve the kubeletplugin support for this feature (abstract from underlying gRPC interface similar to how it's done for the main DRA kubelet interace, better documentation) and only then make the DRA example driver use that improved support?

[kubernetes-prow]

[APPROVALNOTIFIER] This PR is NOT APPROVED

This pull-request has been approved by: harche
Once this PR has been reviewed and has the lgtm label, please assign klueska for approval. For more information see the Code Review Process.

The full list of commands accepted by this bot can be found here.

Details

Needs approval from an approver in each of these files:

• OWNERS

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[harche]

I tested both DRAResourceHealth API generations end-to-end on kind by temporarily redirecting the k8s.io staging modules in go.mod to kubernetes master pseudo-versions (which contain the merged WatchHealthStatus API from kubernetes/kubernetes#139477):

• v1alpha1 path: against a released kindest/node:v1.36.1 kubelet (pre-v1), the health e2e passes via the helper's v1alpha1 compatibility serving — initial Healthy, annotation-forced Unhealthy, and recovery back to Healthy all surface in pod.status.containerStatuses[].allocatedResourcesStatus.
• v1 path: against a kubelet built from kubernetes master (v1.37.0-beta.0.446+4b1379a4a44), the same e2e passes, and the driver logs confirm the kubelet opened the v1 stream (/v1.DRAResourceHealth/NodeWatchResources) rather than falling back to v1alpha1.

The replace block in go.mod is temporary: once WatchHealthStatus ships in a tagged k8s.io/dynamic-resource-allocation release, I'll drop the redirect and bump to the release version.