Updated DRA APIs to v1

[alimaazamat]

fixes #115

[alimaazamat]

/assign

[klueska]

Please see the following for how to gt the DRA driver to work with all of v1, v1alpha1, and v1alpha2: #115 (comment)

[lauralorenz]

I am using the example driver in some kubernetes documentation, and want to queue up a PR that references the tag for the image that will be released with this (in kubernetes/website#51979). Will the new image released be registry.k8s.io/dra-example-driver/dra-example-driver:v0.2.0?

Also for what it's worth, I did run through that tutorial with a local build of the example driver from this branch and it worked great so, I can confirm that it works with k/k 1.34 and the v1 APIs haha 😇

[nojnhuh]

I am using the example driver in some kubernetes documentation, and want to queue up a PR that references the tag for the image that will be released with this (in kubernetes/website#51979). Will the new image released be registry.k8s.io/dra-example-driver/dra-example-driver:v0.2.0?

Yes, I'd like to publish a v0.2.0 image and chart soon after this merges.

Also for what it's worth, I did run through that tutorial with a local build of the example driver from this branch and it worked great so, I can confirm that it works with k/k 1.34 and the v1 APIs haha 😇

Thanks for checking!

[klueska]

I know I saw the changes from NVIDIA/k8s-dra-driver-gpu@920a287 integrated in the PR at some point, but they seem to have been dropped...

[alimaazamat]

I know I saw the changes from NVIDIA/k8s-dra-driver-gpu@920a287 integrated in the PR at some point, but they seem to have been dropped...

I dropped them because I was thinking the multiversion compatability would be handled with runtime-config in kind-cluster-config.yaml?
DRA client is used only for controller components for Nvidia like here https://github.com/NVIDIA/k8s-dra-driver-gpu/blob/main/cmd/compute-domain-controller/main.go#L178
But we only use kubelet plugin which only uses coreclient https://github.com/alimaazamat/dra-example-driver/blob/update-to-v1beta2/cmd/dra-example-kubeletplugin/driver.go#L58

[alimaazamat]

Also having runtime-config: "resource.k8s.io/all=true" causes e2e tests to fail (regardless of including/excluding changes from NVIDIA/k8s-dra-driver-gpu@920a287)

[control-plane-check] kube-apiserver is not healthy after 4m0.00028186s
A control plane component may have crashed or exited when started by the container runtime.
error: error execution phase wait-control-plane: failed while waiting for the control plane to start: kube-apiserver check failed at https://172.18.0.2:6443/livez: client rate limiter Wait returned an error: context deadline exceeded

When excluding changes from NVIDIA/k8s-dra-driver-gpu@920a287 and with runtime-config: "resource.k8s.io/all=v1beta1" we pass

[nojnhuh]

Also having runtime-config: "resource.k8s.io/all=true" causes e2e tests to fail (regardless of including/excluding changes from NVIDIA/k8s-dra-driver-gpu@920a287)

It looks like all only works as api/all, but not for a particular API group. Here we probably have to enable each one individually. Let's enable both v1beta1 and v1beta2 (leaving v1 enabled by default).
https://kubernetes.io/docs/reference/command-line-tools-reference/kube-apiserver/#:~:text=%2D%2Druntime%2Dconfig%20%3Ccomma%2Dseparated%20%27key%3Dvalue%27%20pairs%3E

[alimaazamat]

It looks like all only works as api/all, but not for a particular API group. Here we probably have to enable each one individually. Let's enable both v1beta1 and v1beta2 (leaving v1 enabled by default). https://kubernetes.io/docs/reference/command-line-tools-reference/kube-apiserver/#:~:text=%2D%2Druntime%2Dconfig%20%3Ccomma%2Dseparated%20%27key%3Dvalue%27%20pairs%3E

With runtime-config: "resource.k8s.io/v1beta1=true,resource.k8s.io/v1beta2=true" I tested this with:
1.32 cluster errors out

[api-check] The API server is not healthy after 4m0.001581692s

Unfortunately, an error has occurred:
       context deadline exceeded

This error is likely caused by:
       - The kubelet is not running
       - The kubelet is unhealthy due to a misconfiguration of the node in some way (required cgroups disabled)

1.33 cluster (correctly uses v1beta2)
1.34 cluster (correctly uses v1)

To recap:
1.32 allows v1beta1
1.33 allows v1beta1, v1beta2
1.34 allows v1beta1, v1beta2, v1

So I think defining both v1beta1 and v1beta2 in runtime config doesn't work becasue v1beta2 not available for 1.32
Because with just runtime-config: "resource.k8s.io/v1beta1=true 1.32 cluster works

[nojnhuh]

I ran through all the valid combinations of v1beta1/v1beta2/v1 and 1.32/1.33/1.34 and things are working as I expect, so overall LGTM.

@alimaazamat As soon as the k8s.io dependencies and kind image for the final 1.34.0 release are available let's get a clean CI run before squashing and merging this. Feel free to squash earlier too if that's easier.

[nojnhuh]

/lgtm
/approve

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: alimaazamat, nojnhuh

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

The pull request process is described here

Needs approval from an approver in each of these files:

• OWNERS [nojnhuh]

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