Cannot restart cluster

Ask your workspace administrator to update cluster access rights.

Written by parth.sundarka

Last published at: July 23rd, 2025

Problem

When you attempt to submit a command or query to a cluster, it fails with the following error.

Error: `HTTP Response code: 403, Error message: PERMISSION_DENIED: You do not have permission to autostart <cluster-id>`.


Cause

You do not have permission to restart or start the cluster. 

This error can occur regardless of whether the query submission comes from a notebook, job, the JDBC or ODBC driver, and so on. 


Solution

Ask your workspace administrator to update cluster access rights. They should assign at a minimum the CAN_RESTART permission on the target cluster. 


Alternatively, they can assign CAN_MANAGE. These permissions allow performance of key cluster management actions, including starting, restarting, and terminating clusters. To determine who needs to have permissions assigned for the cluster, refer to the following table.

If the error comes from a …

the entity who needs permission is… 

Notebook

you.

Query triggered through JDBC/ODBC

the user identity used to authenticate to Databricks.

Job

the user mentioned in the job’s run_as parameter.


For more information, your workspace admin can refer to the “Configure compute permissions” section of the Manage compute (AWS | Azure | GCP) documentation.


You can instead ask someone with CAN_RESTART or CAN_MANAGE permissions to start the cluster, and then try to run your command again. For this option, ensure you have the CAN_ATTACH_TO privilege.