feat: Add successpolicy

[gaocegege]

SuccessPolicy is used in both PyTorchJob and TFJob. Thus I propose to add it in common.

Signed-off-by: cegao [email protected]

[gaocegege]

/assign @terrytangyuan @Jeffwan @zw0610

[terrytangyuan]

This is great. Although I wonder if we can use this opportunity to re-think the approach. Instead of enums, would a label selector similar to what Argo Workflows is doing here be more flexible and allows more fine-grained control over the success conditions/policies? https://github.com/argoproj/argo-workflows/blob/master/examples/k8s-kubeflow-jobs.yaml#L24-L25

[gaocegege]

It is used in Katib. I think it works, but I think we should support successPolicy to keep API consistency.

cc @andreyvelich

[andreyvelich]

Yes, we are using successCondition with GSON format in our APIs to define condition for Katib Trial's Workers, similar to Argo Workflows.
Probably, we can think how to use it in Training Operators.

[terrytangyuan]

Let's use kubeflow/training-operator#1507 to track and discuss separately.

[terrytangyuan]

This is LGTM for consistency. Others PTAL.

[google-oss-prow]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: terrytangyuan

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 [terrytangyuan]

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

[gaocegege]

/assign @zw0610 @Jeffwan

[zw0610]

LGTM. Meanwhile, could you add descriptions for SchedulingPolicy and SuccessPolicyAllWorkers to explain the expected behavior?

[gaocegege]

SGTM

[gaocegege]

/hold