• Verified DB Versions: Databricks SQL 2026.39
• Verified Client Version: Flyway v13 & driver v3.4.2
• Maintainer: Redgate

Supported Versions and Support Levels

• For information regarding the supported version and support levels available, please see Supported Databases for Flyway
• For information regarding the Flyway features available, please see Flyway feature summary

Driver

Copy the host and HTTP path from the Connection details tab of your SQL warehouse, for example:

jdbc:databricks://host:443/default;transportMode=http;ssl=1;httpPath=/sql/1.0/warehouses/warehouse-id;EnableArrow=0;AuthMech=3;ConnCatalog=catalog

• ConnCatalog is required. Set it in the URL or jdbcProperties.
• Set schemas and defaultSchema to schema names within that catalog. If neither is set, Flyway uses the schema in the URL path (default above) and creates the schema history table there.
• Flyway sets UserAgentEntry=Redgate_Flyway unless you set UserAgentEntry in jdbcProperties.

Java Usage

Databricks support is a separate dependency for Flyway and will need to be added to your Java project to access these features.

Maven

Redgate

<dependency>
    <groupId>com.redgate.flyway</groupId>
    <artifactId>flyway-database-databricks</artifactId>
</dependency>

Open Source

<dependency>
    <groupId>org.flywaydb</groupId>
    <artifactId>flyway-database-databricks</artifactId>
</dependency>

Gradle

Redgate

buildscript {
    dependencies {
        implementation "com.redgate.flyway:flyway-database-databricks"
    }
}

Open Source

buildscript {
    dependencies {
        implementation "org.flywaydb:flyway-database-databricks"
    }
}

SQL Script Syntax

• Databricks SQL syntax with statement delimiter ;
• Identifiers are quoted with backticks, for example `my schema`
• BEGIN ... END compound statements, such as procedure bodies

Compatibility

• Any SQL script executed by Flyway can be run in the Databricks SQL editor (after the placeholders have been replaced)

Example

/* Single line comment */
CREATE TABLE test_data (
  id BIGINT GENERATED ALWAYS AS IDENTITY,
  value STRING NOT NULL
);

/*
Multi-line
comment
*/

-- Sql-style comment

CREATE OR REPLACE PROCEDURE add_test_data(new_value STRING)
LANGUAGE SQL
SQL SECURITY INVOKER
AS
BEGIN
  INSERT INTO test_data (value) VALUES (new_value);
END;

-- Placeholder
INSERT INTO ${tableName} (name) VALUES ('Mr. T');

Authentication

Flyway passes your URL and jdbcProperties to the Databricks JDBC driver, which handles authentication. Choose a method based on where Flyway runs:

With a personal access token, set password to the token. With OAuth machine-to-machine, set the OAuth2ClientId and OAuth2Secret JDBC properties to the service principal's application ID and a Databricks OAuth secret. With OAuth user-to-machine, the driver opens a browser to sign in and caches the tokens, encrypted with the passphrase in the TokenCachePassPhrase JDBC property.

For the full list of driver authentication options, refer to Databricks' documentation.

For how to use authentication methods, see Connecting to environments. For credential storage and retrieval, see Storing and retrieving credentials.

Notes

For Databricks Lakebase, which Flyway connects to through its PostgreSQL support, see Databricks Lakebase.

Limitations

• Due to a change in the Java 16+ runtime which Flyway depends upon, a formatting error will occur in the Databricks JDBC driver unless you define the handling of such, that can be done by either:
  • appending EnableArrow=0 to your JDBC connection string
  • adding --add-opens java.base/java.lang=ALL-UNNAMED. This can be done via the command line or environment variables
• Databricks doesn't support transactional DDL, so Flyway runs migrations on Databricks without a transaction and each statement takes effect as soon as it runs. If a statement fails partway through a migration, the statements before it stay applied. Flyway records the migration as failed: you must undo the applied statements by hand, then run repair before migrating again. Keeping each migration to a single DDL statement limits the impact of a failure.
• Parallel migrations as described here are unavailable in Databricks. Don't run more than one Flyway instance against the same schema history table at once.
• Clean doesn't remove procedures or volumes