Databricks

Overview

The Databricks connector runs SQL through the Statement Execution API and starts jobs through the Jobs API. Results come back as typed rows with the column schema attached, and every query in a run shows its statement ID and duration in the trace.

Data teams use it to start a reconciliation workflow when a nightly job finishes, to page on-call in Slack when a job run fails and to load scored rows from Kafka-fed tables into Postgres.

Supported workspaces

• Databricks on AWS, Azure and Google Cloud

• Serverless and pro SQL warehouses

• Unity Catalog tables, with three-part names in queries

How queries work

Tessel submits each statement to the warehouse you pick and waits up to 30 seconds inline. Longer queries continue in the background and Tessel polls for the result. If the warehouse is stopped, the first query starts it and the trace shows the wait.

POST /api/2.0/sql/statements
{
  "warehouse_id": "4b9b953939869799",
  "statement": "SELECT * FROM finance.recon.open_items WHERE run_date = :d",
  "parameters": [{ "name": "d", "value": "2026-10-06", "type": "DATE" }],
  "wait_timeout": "30s",
  "disposition": "EXTERNAL_LINKS"
}

POST /api/2.0/sql/statements
{
  "warehouse_id": "4b9b953939869799",
  "statement": "SELECT * FROM finance.recon.open_items WHERE run_date = :d",
  "parameters": [{ "name": "d", "value": "2026-10-06", "type": "DATE" }],
  "wait_timeout": "30s",
  "disposition": "EXTERNAL_LINKS"
}

POST /api/2.0/sql/statements
{
  "warehouse_id": "4b9b953939869799",
  "statement": "SELECT * FROM finance.recon.open_items WHERE run_date = :d",
  "parameters": [{ "name": "d", "value": "2026-10-06", "type": "DATE" }],
  "wait_timeout": "30s",
  "disposition": "EXTERNAL_LINKS"
}

Large results

Results over 25 MB are fetched as chunks through external links and streamed into the next step. Write-backs go through MERGE on a key you choose, so a retried batch never writes rows twice.

Permissions needed

Create a service principal for Tessel and connect with OAuth machine-to-machine. It needs use access to one SQL warehouse and read access to the tables you query. Write access is only required for the staging schema.

• CAN USE on the SQL warehouse

• USE CATALOG, USE SCHEMA and SELECT on source tables

• MODIFY on the staging schema and CAN MANAGE RUN on jobs you trigger