How can I run a pipeline when a file is uploaded to a GCP bucket (Kubeflow 1.3 multi-user)?

Question

I get the following error:

"Failed to authorize with API resource references: Failed to authorize with API resource references: PermissionDenied: User 'myServiceAccount@appspot.gserviceaccount.com' is not authorized with reason: (request: ResourceAttributes{Namespace:namespace1,Verb:list,Group:pipelines.kubeflow.org,Version:v1beta1,Resource:experiments,Subresource:,Name:,

Code

def main(data, context):
  client = kfp.Client(host=HOST, client_id=CLIENT_ID,namespace=NAMESPACE)
  logging.info(client.list_experiments(namespace=NAMESPACE))

Have you gone through the [pipeline authentication](https://www.kubeflow.org/docs/distributions/gke/pipelines/authentication-pipelines/) process? — JMA, Sep 30 '21 at 10:00
@JMA Thanks for answering. Yes, I have. The problem was that when I set the correct service account, a warning message appeared (see my answer below). That was confusing — pablofiumara, Sep 30 '21 at 15:54

score 1 · Answer 1 · answered Sep 30 '21 at 15:53

Now my GCP function is able to list pipelines.

I had to use the following service account: KF-NAME-user@KF-PROJECT.iam.gserviceaccount.com .

This message appears in the logs but it's just a warning message No such file or directory: '/var/run/secrets/kubernetes.io/serviceaccount/namespace'

How can I run a pipeline when a file is uploaded to a GCP bucket (Kubeflow 1.3 multi-user)?

1 Answers1