Databricks Integration

Connect Databricks to enable the AI agent to query data in SQL Warehouses, monitor ETL job runs, and check cluster health during investigations.

Capabilities

Once connected, the AI agent can:

Capability	Description
SQL Queries	Execute read-only SQL queries against SQL Warehouses
Job Monitoring	Monitor ETL and data pipeline job runs
Cluster Health	Check Databricks cluster status and health
Data Exploration	Browse schemas, tables, and data catalogs

note

Read-only access: This integration only executes SELECT queries. UPDATE, DELETE, INSERT, and other data-modifying commands are blocked at the query level.

Prerequisites

• A Databricks workspace with API access
• A Personal Access Token (PAT) or OAuth token
• At least read permissions on the data you want to query

Setup

1

Get Your Access Token

1. Log in to your Databricks workspace
2. Click your profile icon in the top-right corner
3. Go to User Settings → Developer → Access tokens
4. Click Generate new token
5. Give it a description (e.g., "Autoheal Integration") and set an expiration
6. Copy the generated token

2

Add Integration in Autoheal

1. Go to Integrations in Autoheal
2. Click Databricks
3. Enter a name (e.g., "Production Databricks")

3

Configure Credentials

Enter the following:

• Workspace URL: Your Databricks workspace URL (e.g., https://adb-1234567890.7.azuredatabricks.net or https://your-workspace.cloud.databricks.com)
• Access Token: Your Personal Access Token (starts with dapi...)

4

Test and Save

Click Test Connection to verify, then Save.

Required Permissions

The access token should have at least these permissions:

Permission	Why It's Needed
CAN_USE on SQL Warehouse	Execute SQL queries
CAN_VIEW on Jobs	Monitor job runs
CAN_ATTACH_TO on Clusters	Check cluster status
SELECT on Tables/Views	Read data from tables

tip

Create a dedicated service principal or machine user for Autoheal with only read permissions. Avoid using personal tokens with admin access.

Example Queries

Once connected, you can ask the AI agent questions like:

Query the orders table for failed transactions in the last hour

Show me the status of the nightly ETL job

Which Databricks clusters are currently running?

List the top 10 error messages from the application_logs table

Troubleshooting

401 Unauthorized Error

• Verify the access token is correct and has not expired
• Check that the token has the required permissions
• Ensure the workspace URL matches where the token was created

No SQL Warehouse Available

• Verify at least one SQL Warehouse is running in your workspace
• Check that the token has CAN_USE permission on the warehouse
• Ensure the warehouse is not paused or stopped

Connection Timeout

• Check that the workspace URL is correct and accessible
• Verify network connectivity to Databricks APIs
• Ensure there are no firewall rules blocking access