The mlflow module now supports the scenario 5. This can be achieved by putting the artifact_Proxied_Access variable in the locals file to true.
MLflow pods on GKE now use Workload identity to enable the mlflow-tracking kubernetes service account to use the Terraform GKE SA service account to access the Cloud Storage.
The gcp-minimal and gcp-kubeflow-kservehave reverted to using thegkemodule instead of thegoogle_container_cluster` resource.
artifact_Proxied_Accessvariable in the locals file to true.mlflow-trackingkubernetes service account to use theTerraform GKE SAservice account to access the Cloud Storage.gcp-minimaland gcp-kubeflow-kservehave reverted to using thegkemodule instead of thegoogle_container_cluster` resource.READMEis now updated.