This is expected behavior with mlops-stacks and not an issue with your Terraform version or the CLI.

The main problem is that your Azure Databricks workspace does not have Unity Catalog enabled or assigned. The mlops-stacks templates assume Unity Catalog by default. Because of that, Terraform tries to set permissions on jobs and MLflow experiments and also tries to create a registered model, all of which require a Unity Catalog metastore. When the workspace doesn’t have a metastore, Databricks returns errors like “ACLs for job / mlflowExperiment are disabled or not available in this tier” and “No metastore assigned for the current workspace”.

Once you create and attach a Unity Catalog metastore to the workspace, these errors go away and the bundle deploy works as expected. After assigning the metastore, make sure you also set a default catalog and schema and that your user or service principal has the required privileges.

If you just want to get things running quickly in a dev setup without Unity Catalog, you can temporarily remove the permissions blocks from the job and MLflow experiment resources and skip registered model creation, but this is only a workaround. The official and supported path for mlops-stacks is with Unity Catalog enabled.

For CI/CD on Azure Databricks with GitLab, the recommended approach is to use a service principal (not a user PAT) and Databricks Bundles. The Databricks Bundles CI/CD documentation applies directly to GitLab, and the mlops-stacks repo has workflow examples you can easily translate from GitHub Actions to GitLab CI.

Docs that helped me:

• Unity Catalog setup: https://docs.databricks.com/data-governance/unity-catalog/get-started.html

• MLflow + Unity Catalog: https://docs.databricks.com/mlflow/models.html

• Bundles CI/CD: https://docs.databricks.com/dev-tools/bundles/ci-cd.html

In short: enable Unity Catalog, redeploy the bundle, and the errors you’re seeing should be resolved.