SQL warehouse will not launch after upgrading to serverless

Remove the overriding runtime version from the SQL warehouse configuration.

Written by kingshuk.das

Last published at: June 17th, 2025

Problem

When attempting to start a SQL warehouse after upgrading it from pro to serverless, you encounter an error that prevents the warehouse from launching. 

"Clusters are failing to launch. Cluster launch will be retried. Request to create a cluster failed with an exception: Serverless warehouse cannot start since warehouse `<warehouse-id>` has runtime version overrides. Please remove the overrides to continue using the warehouse."

 

Cause

You’re using a custom Databricks runtime version with your SQL warehouse. When you change the warehouse from pro to serverless, any existing runtime version overrides become incompatible with the new serverless configuration. 

 

Serverless SQL warehouses do not support runtime version overrides, resulting in the error.

 

Solution

Important

You may want to set test_overrides to NULL to solve this problem. NULL overrides all custom Apache Spark configurations for the warehouse. The following solution shows you how to indicate specific Spark configurations using key-value pairs.

 

 

  1. In your SQL warehouse configuration, indicate the runtime version overrides to remove as a key-value pair within the spark_conf field. The following code provides an example.
{
    "confs":
{
    "test_overrides": {
        "cluster_attributes": {
            "spark_conf": {
                key: <key>
                value: <value>
                          }
                              }
                      }
    }
}

 

  1. Use the API to execute a POST call with this updated configuration. Refer to the Update a warehouse (AWS | Azure | GCP) API documentation for details. 
  2. After updating the configuration, restart the SQL warehouse. The warehouse automatically takes the default Databricks Runtime upon restarting.