Reverse ETL with Databricks

Reverse ETL is the pattern of reading data out of Databricks and writing it into another destination — a SaaS app (Salesforce, HubSpot, …), another database, a file, or a queue.

Build a Reverse ETL flow

Treat the Databricks connection as a regular database source. Etlworks executes a JDBC SELECT (or a SQL query you provide) against the Databricks compute resource, optionally transforms the result, and writes it to the chosen destination.

1. Create a Databricks connection.
2. Create the destination connection (SaaS app, database, file storage, queue, …).
3. In Flows, create the right flow type for the destination — for example, Database to Database if writing to another RDBMS, or Database to file for an export.
4. Add a transformation. Set FROM to the Databricks connection and the source table or query. Set TO to the destination connection and object.
5. Optional: enter a Source SQL on the transformation to control the exact query Databricks runs (joins, filters, aggregations, time-bounded windows).
6. Optional: enable an incremental read with change replication using High Watermark based on a Delta column such as __processed_at or a monotonic sequence.
7. Save and schedule the flow.

Note: Etlworks reads from Databricks via JDBC. Delta Sharing and Databricks SQL's predictive features are not used by the reverse-ETL path today — the read is a standard SELECT executed by the Databricks compute resource.