Add Job interface to restrain the behaviors of different Job implementations

[kerthcet]

What would you like to be added:

Currently, we have a default integrated implementation of batchv1 job, we achieve the expected behaviors by the experiences of kubernetes, and most importantly, we know the project well.

But for people who wants to implement a special controller in kueue, they have little knowledges and they may don't know where to start. Documents are great and always needed, but they're not binding.

So we should make a contract known as interface in golang, which defines the behaviors of the various jobs. E.g.:

type Job interface {
	Suspend()
	UnSuspend()
	QueueName()
	IsSuspend() bool
	Start() error
	Stop() error
	ConstructWorkload() (*kueue.Workload, error)
}

And we can wrap batch job like below, BatchJob is an implement of kueue.Job

type BatchJob struct {
	job batchv1.Job
}

Why is this needed:

Completion requirements:

This enhancement requires the following artifacts:

• Design doc
• API change
• Docs update

The artifacts should be linked in subsequent comments.

[tenzen-y]

That sounds great!
The job interface helps support Kueue with custom jobs such as TFJob, PytorchJob, and MPIJob in kubeflow/training-operator and kubeflow/mpi-operator/v2.

[alculquicondor]

I would also add some form of InjectNodeAffinity

[tenzen-y]

I would also add some form of InjectNodeAffinity

Does this mean the following?

kueue/pkg/controller/workload/job/job_controller.go

Lines 263 to 278 in 6e403b5

	nodeSelector, err := r.getNodeSelectors(ctx, w)
	if err != nil {
	return err
	}
	if len(nodeSelector) != 0 {
	if job.Spec.Template.Spec.NodeSelector == nil {
	job.Spec.Template.Spec.NodeSelector = nodeSelector
	} else {
	for k, v := range nodeSelector {
	job.Spec.Template.Spec.NodeSelector[k] = v
	}
	}
	} else {
	log.V(3).Info("no nodeSelectors to inject")
	}

[alculquicondor]

yes

[tenzen-y]

SGTM

[k8s-triage-robot]

The Kubernetes project currently lacks enough contributors to adequately respond to all issues and PRs.

This bot triages issues and PRs according to the following rules:

• After 90d of inactivity, lifecycle/stale is applied
• After 30d of inactivity since lifecycle/stale was applied, lifecycle/rotten is applied
• After 30d of inactivity since lifecycle/rotten was applied, the issue is closed

You can:

• Mark this issue or PR as fresh with /remove-lifecycle stale
• Mark this issue or PR as rotten with /lifecycle rotten
• Close this issue or PR with /close
• Offer to help out with Issue Triage

Please send feedback to sig-contributor-experience at kubernetes/community.

/lifecycle stale

[tenzen-y]

/lifecycle frozen

[kerthcet]

We have a plan to integrate with other workloads, also see #499, I'd like to write a design doc at first.
/assign

[tenzen-y]

I'm also working on the kubeflow project.

So I'm thinking of giving feedback on the design doc since I'd like to support Kueue with training-operator and mpi-operator v2.

[kerthcet]

Thanks @tenzen-y , when you're ready, plz also cc me, thanks.

[alculquicondor]

From the messages above, I'm not sure who is writing the design doc. It sounds like @kerthcet, is that correct?

[tenzen-y]

From the messages above, I'm not sure who is writing the design doc. It sounds like @kerthcet, is that correct?

Probably, @kerthcet is writing the doc.

[tenzen-y]

If @kerthcet doesn't have enough time, I can write it.

[kerthcet]

Yes, I'm writing now, almost finished.

[tenzen-y]

Sounds great.

[alculquicondor]

@kerthcet do you have a draft to share? It might be beneficial to provide you feedback already.

[alculquicondor]

Oh, from slack, it seems that Kante is OOO until the end of the month

[tenzen-y]

Thank you for letting us know.

[kerthcet]

FYI #537 KEP is ready for first round of review.

[alculquicondor]

This is completed

/close

[k8s-ci-robot]

@alculquicondor: Closing this issue.

In response to this:

This is completed

/close

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository.