[sdk] gcp secret depreciated

[N214]

Environment

• KFP version: 1.7

• KFP SDK version: 1.8.22

• All dependencies version:

Steps to reproduce

In, kfp v1, I use the line below to pass the service account to the pipeline pod

dsl.get_pipeline_conf().add_op_transformer(gcp.use_gcp_secret('user-gcp-sa'))

However, by using kfp.v2.dsl, I get

module 'kfp.v2.dsl' has no attribute 'get_pipeline_conf'

Using the new method below works

@dsl.pipeline(
    name='GCP access in pipeline',
)
def cred_pipeline():
    from kfp.gcp import use_gcp_secret
    container_op = read_gcs().apply(use_gcp_secret('user-gcp-sa'))

But I see that the gcp file has been moved to kfp.depreciated in v2, do we have any alternative to use a service account key in v2?

Expected result

Materials and Reference

---

Impacted by this bug? Give it a 👍.

[zijianjoy]

@N214 For running Pipelines on GCP, we are recommending to use Vertex SDK, where you can specify the service account there.

cc @connor-mccarthy

[N214]

@N214 For running Pipelines on GCP, we are recommending to use Vertex SDK, where you can specify the service account there.

Ah interesting ! Thank you

Just to explain a little bit more my use case here, I have on-prem hardware that I wish to use to train my models, but some datasets are store in GCS or Bigquery, so I just need a service account to be passed to the pipeline pod to have access to GCP. The pipelines should run on prem, not on GCP.

Thank you

[github-actions[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[github-actions[bot]]

This issue has been automatically closed because it has not had recent activity. Please comment "/reopen" to reopen it.