Hi Team,

We are investigating an issue where Databricks worker instances running on AWS are not registering with AWS Systems Manager (SSM).

Environment

• Databricks Runtime: 17.3 LTS
• Worker OS: Ubuntu 24.04
• AWS EC2 worker instances launched by Databricks
• IAM instance profile attached with the required SSM permissions (including AmazonSSMManagedInstanceCore)

Validation Performed

• Verified the EC2 instances launch successfully.
• Confirmed the correct IAM instance profile is attached.
• Verified required SSM permissions are present.
• Validated outbound network connectivity.
• Verified security groups and routing configuration.

Issue Observed

The Databricks worker EC2 instances do not register as Managed Nodes in AWS Systems Manager.

During troubleshooting we observed:

• The amazon-ssm-agent binary is not available on the worker nodes.
• No SSM Agent service is running.
• Session Manager status remains Not Connected.
• The instances do not appear in the AWS Systems Manager Managed Nodes inventory.

Troubleshooting Performed

• Created a test Databricks compute with the required instance profile.
• Attempted to install and start the Amazon SSM Agent using a cluster-scoped init script.
• Verified the init script execution completed successfully.
• Despite the installation attempt, the EC2 instances still do not register with AWS Systems Manager.

Questions

Could you please help clarify the following?

1. Are Databricks Runtime 17.3 Ubuntu worker images expected to include the Amazon SSM Agent by default?
2. Is installing the Amazon SSM Agent through cluster-scoped init scripts a supported approach?
3. What is the recommended method for enabling AWS Systems Manager access on Databricks-managed EC2 worker nodes?
4. Are there any Databricks restrictions or known limitations that would prevent SSM Agent registration?

Any guidance or recommended best practices would be appreciated.

Thank you