feat: Add pipeline run parallelism config

[sduvvuri1603]

Summary

• add pipeline_run_parallelism to dsl.PipelineConfig (proto, Go, Python) and thread it through the compiler/runtime so the value lands in the workflow bundle and Kubernetes platform spec
• persist the limit in kfp-pipeline-config and apply it to Argo spec.parallelism; extend backend tests and regenerate proto goldens
• update docs and samples to account for the new config

What pipeline_run_parallelism does

pipeline_run_parallelism lets authors cap the number of tasks that can execute concurrently in a pipeline run. When set, the value is stored in kfp-pipeline-config and the compiler writes it to the workflow manifest (spec.parallelism) so Argo enforces the limit.

Validation

• SDK compilation golden (pipeline_with_run_parallelism.yaml) shows pipelineRunParallelism in the compiled spec
• Backend compiler golden updates confirm spec.parallelism is populated
• TestCreateRun_PipelineRunParallelismConfigMap exercises the config map persistence path
• Manual before/after runs verify the config map entry and resulting workflow manifest

[google-oss-prow]

[APPROVALNOTIFIER] This PR is NOT APPROVED

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

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

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

[google-oss-prow]

Hi @sduvvuri1603. Thanks for your PR.

I'm waiting for a kubeflow member to verify that this patch is reasonable to test. If it is, they should reply with /ok-to-test on its own line. Until that is done, I will not automatically test new commits in this PR, but the usual testing commands by org members will still work. Regular contributors should join the org to skip this step.

Once the patch is verified, the new status will be reflected by the ok-to-test label.

I understand the commands that are listed here.

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.

[alyssacgoins]

/retest

[google-oss-prow]

@sduvvuri1603: Cannot trigger testing until a trusted user reviews the PR and leaves an /ok-to-test message.

In response to this:

/retest

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.

[hbelmiro]

/ok-to-test

[hbelmiro]

/retest

[nsingla]

is this intentional to hardcode the namespace, what if I deploy with a different namespace, is this getting replaced/overriden somewhere?

[sduvvuri1603]

No, it was not intentional.
Thanks for pointing that out! I’ll change this binding to use a namespace variable($(kfp-namespace)) so it follows whatever namespace users deploy into.

[nsingla]

I think we need a pipeline in test_data/valid/*compiled/essential that overrides this value.

[VaniHaripriya]

It may be more appropriate to add this entry to the CHANGELOG.

[sduvvuri1603]

Sure, but will this be a part of a new section called "Unreleased Features" ? because I only see version release details in the file.

[VaniHaripriya]

I believe the PR will be included here as part of the release process. @mprahl , could you confirm if that’s correct?

[nsingla]

is this required given https://github.com/kubeflow/pipelines/pull/12442/files#diff-631f096829954a5226e87a37347ab04d3e551465fdf0371b37fd9f9363c0c156R120 will set it to None if value is None?

[sduvvuri1603]

Yes, it's right about the serialization part. But we need this guard to prevent a crash during initialization. Since init passes None by default, removing this check would cause it to hit the isinstance line and fail immediately. This just ensures we can safely create the object with no value set.

[nsingla]

I guess then you can just add 1 if statment:

        if value:
          if not isinstance(value, int):
              raise ValueError(
                  'pipeline_run_parallelism must be an integer if specified.')
          if value <= 0:
              raise ValueError(
                  'pipeline_run_parallelism must be a positive integer.')
       self._pipeline_run_parallelism = value

[nsingla]

can we actually set this to None here if we have an explicit test to test +ve values?

[sduvvuri1603]

Yep, That makes sense. Will set this to none

[nsingla]

@sduvvuri1603 can you please add what this config is suppose to do, to the PR description? and a section about how you;ve validated the functionality.

[nsingla]

Isn't 7 too high when the number of tasks in this pipeline is just 1? May be you should add more components to it or add a parallelFor loop and iterate over > pipeline_run_parallelism constants, so that we can validate that the config actually works.
Also what validation logic did you add to confirm the number of tasks created for a pipeline with this config?

[sduvvuri1603]

This specific test case is part of the SDK compilation suite to verify that the pipeline_run_parallelism field is correctly serialized from the Python SDK into the compiled YAML's PlatformSpec. It relies on the 'Golden File' comparison for validation here (ensuring the YAML contains pipelineRunParallelism: 7 correctly populated)

(It is not related to actual runtime limit covered by the backend integration tests where we submit these workflows to Argo is my understanding) Pls Lmk if this is correct!

[nsingla]

Any pipeline yaml file in this directory will be part of the end to end tests, so yes, the workflow will get submitted to argo.