MLFlow Remote model registry connection is not working in Databricks

[its-kumar]

Dear community,

I am having multiple Databricks workspaces in my azure subscription, and I have one central workspace. I want to use the central workspace for model registry and experiments tracking from the multiple other workspaces.

So, If I am training and registering some model from any of the workspaces, it should register inside my central workspace.

I am following the following notebook for the same https://docs.databricks.com/_extras/notebooks/source/mlflow/mlflow-model-registry-multi-workspace.ht..., but I am getting this error whenever I create register model.

MlflowException: API request to endpoint /api/2.0/mlflow/runs/create failed with error code 404 != 200. Response body: ''

I am not able to find any solution. Please guide me.

Thanks in advance.

[User16752242622]

Hello, This looks like a bug with the command MlflowClient().get_latest_versions("SOME_REGISTERED_MODEL")

Please check this: https://github.com/mlflow/mlflow/issues/5171

[Anonymous]

@Kumar Shanu​ :

The error you are seeing (API request to endpoint /api/2.0/mlflow/runs/create failed with error code 404 != 200) suggests that the API endpoint you are trying to access is not found. This could be due to several reasons, such as incorrect authentication, wrong endpoint URL, or missing permissions.

To troubleshoot this issue, you can try the following steps:

1. Double-check that you have correctly set up the connection to the central workspace in the MLFLOW_TRACKING_URI environment variable. The URI should be in the format databricks://<profileName>@<workspaceUrl>
2. Ensure that the central workspace has the necessary permissions to create runs and register models from the other workspaces. You can check this by going to the workspace settings in the Databricks UI and verifying that the relevant permissions are enabled.
3. Verify that the API endpoint /api/2.0/mlflow/runs/create exists in the central workspace by checking the API documentation or using a REST client to make a test request.
4. Check the Databricks workspace logs to see if there are any additional error messages or clues as to why the API request is failing.