How to deploy GKE workload that uses aiplatform.models.get in the containerized application

18 views Asked by At

I've been trying to move my application from GCP cloud run to GKE. I am using a finetuned language model trained using vertex ai and the workload when deployed on the cluster keeps throwing crashloop error. When I go through the logs, I can see the service account doesn't have permission to access aiplatform.models.get.

I've provided all the required permission to this service account used by the cluster and it still persist. Error:

metadata {
  key: "permission"
  value: "aiplatform.models.get"
}
]
[2024-02-09 19:06:54 +0000] [7] [INFO] Worker exiting (pid: 7)

I've updated the permission for the service account using UI as well as this command:

gcloud projects add-iam-policy-binding <PROJECT_ID>  --member=serviceAccount:<service-account-name>@<PROJECT_ID>.iam.gserviceaccount.com --role=roles/aiplatform.viewer

note: I've tried adding aiplatform.admin and editor role but the issue continue to persist.

0

There are 0 answers