Skip to main content
Version: 6.1.0
Databricks is a unified analytics platform built on Apache Spark, providing data engineering, data science, and machine learning capabilities in the cloud. Use the Python Connector for SQL warehouses and clusters.
Installation
Required packages: apache-superset[databricks]
pip install apache-superset[databricks]
Drivers
PyPI Package: databricks-sql-connector
databricks://token:{access_token}@{host}:{port}?http_path={http_path}&catalog={catalog}&schema={schema}
Supported Features
JOINsSubqueriesDynamic SchemaCatalog SupportDynamic CatalogSSH TunnelingQuery CancellationFile UploadUser ImpersonationCost EstimationSQL Validation
Feature Score: 70/201
Time Grains
Common Time Grains:
SECONDMINUTEHOURDAYWEEKMONTHQUARTERYEAR
Extended Time Grains:
FIVE_SECONDSTHIRTY_SECONDSFIVE_MINUTESTEN_MINUTESFIFTEEN_MINUTESTHIRTY_MINUTESHALF_HOURSIX_HOURSWEEK_STARTING_SUNDAYWEEK_STARTING_MONDAYWEEK_ENDING_SATURDAYWEEK_ENDING_SUNDAYQUARTER_YEAR
Help improve this documentation by editing the engine spec:

Upgrading legacy connections​

Workspaces upgrading from 6.0 with a legacy Databricks connection have their databricks+connector:// URI automatically migrated to databricks:// during superset db upgrade. The migration preserves credentials, HTTP path, catalog, and the rest of the URI. No API access or manual connection recreation is needed.

Edit the connection to review the saved host, HTTP path, catalog, schema, and masked access token, or enter a replacement token. When creating a connection with a SQLAlchemy URI, Connect this database using the dynamic form instead opens the field-based form.

Downgrading restores the old scheme on locally converted connections only when their URI, settings, and modification timestamp still match the migrated state. Imported connections carrying the conversion marker are also eligible for rollback. Connections created natively with databricks:// are not reverted. Malformed extra settings do not block the upgrade: their text is preserved, and only the URI scheme changes. The migration keeps fingerprints in a bookkeeping table so these connections can also be safely downgraded without parsing their settings. Invalid settings may still need correction before the connection can execute queries.

The Default Catalog field is optional. Connections converted from the legacy connector retain the legacy runtime default catalog when none is configured, so existing dataset catalogs and schema permissions remain consistent. Clearing a field removes its corresponding legacy connect_args override. Changing the connection destination requires re-entering the access token.

The new-connection picker offers Databricks, not Databricks (legacy). Existing legacy connections can still be edited with their saved form.

To create a separate connection after editing one, close the edit dialog and choose + Database. The database picker starts a new connection without reusing the edited connection's name or URI, even if the previous connection was still loading when the edit dialog was closed.