[FEATURE] Create way to enable/disable model serving via Terraform

[arpitjasa-db]

Unclear how to enable serving via Terraform. Might not be possible? Possible solution includes adding a flag to enable serving at instantiation, but may involve use of private APIs.

Databricks docs: https://docs.databricks.com/applications/mlflow/model-serving.html MLflow model Terraform resource: https://registry.terraform.io/providers/databrickslabs/databricks/latest/docs/resources/mlflow_model

[arpitjasa-db]

Create a mlflow_model_serving resource with expected behavior as follows: Input Vars

• registered_model_name or experiment_id

Behavior

• Resource creation enables serving on the model via POST request at /mlflow/endpoints/enable
• Resource deletion disables serving on the model via POST request at /mlflow/endpoints/disable
• Resource modification (changing input var) disables serving on the old model and enables serving on the new model (see force_replace attribute)

It would live in this folder: https://github.com/databrickslabs/terraform-provider-databricks/tree/master/mlflow and can use https://github.com/databrickslabs/terraform-provider-databricks/blob/master/mlflow/resource_model.go as an example

[nfx]

@arpitjasa-db and how would we know that endpoint is running or has the configuration that we told it to have?

If you're up for picking it up in your team - we'll guide you. It may be an interesting project for Vladimir.

[smurching]

@jensendw I was just chatting with @nfx who mentioned you might be interested in helping contribute this functionality. That's great! Are you still interested/would you be open to discussing it some time?

[nkvuong]

Close this because of #2054