Assigning Dedicated (SINGLE_USER) ML Clusters to a Group in Databricks

[Mr_7199]

I'm working with Databricks Runtime ML and have configured a cluster in Dedicated access mode (formerly SINGLE_USER). The documentation indicates that a compute resource with Dedicated access can be assigned to a group, allowing user permissions to automatically down-scope to the group’s permissions.

However, when I assign the cluster to a group using the API/CLI, group members can see the cluster in the workspace but are unable to attach it to a notebook. Has anyone successfully configured a dedicated ML cluster to be attached by multiple users via group assignment? Are there any specific procedures or workarounds or additional permission assignments post-creation) that enable this functionality?

For reference, please see the https://docs.databricks.com/aws/en/machine-learning/databricks-runtime-ml

[Isi]

Hey @Mr_7199 

Yes, I’ve successfully configured a dedicated ML cluster assigned to a group.

Here are three things to check:
1.Cluster Policy – Ensure the cluster policy does not impose restrictions. Using an unrestricted policy simplifies testing.
2.Permissions – Verify that users in the group (add the group directly to the permissions) have the CAN ATTACH TO permission on the cluster and CAN USE if a policy applies.
3.Group-Based Execution – When a cluster is assigned to a group (currently in public preview), jobs and notebooks execute under the group’s identity rather than individual users. Ensure the group has the necessary permissions to attach and run workloads on the cluster.

Let me know if you need further details!:)

Isi