Data & Docs · CONNECTIONRun Databricks jobs and SQL statements, read runs, outputs, warehouses, clusters and catalogs, and start clusters (Databricks on AWS workspaces).
Explore 16 workflow actions for Databricks, including account permissions and operation limits.
Choose an action in the workflow builder, connect an account, and map its inputs. These actions are available in the catalog; your account access and a successful test determine whether they work for your setup.
Ask Databricks to cancel a job or task run; it stops asynchronously.
A repeated request may repeat the external action. Review the previous result before retrying a write.
Ask Databricks to cancel a running statement.
A repeated request may repeat the external action. Review the previous result before retrying a write.
Read a cluster's configuration and state.
Read the identity the token belongs to.
Read a job's settings.
Read a run's state, tasks and timing.
Read a task run's output, such as a notebook's exit value (first 5 MB).
Read a statement's status, manifest and first result chunk.
List the Unity Catalog catalogs the token can see.
List the workspace's clusters.
List jobs, optionally by exact name.
List runs, newest first, optionally for one job and only active or completed ones.
List the SQL warehouses the token can use.
Trigger a job run now, optionally with job parameters or only some tasks. NodeHub's step key is the run's idempotency token, so a retry never starts a second run.
A repeated request may repeat the external action. Review the previous result before retrying a write.
Run a SQL statement on a SQL warehouse and wait up to the timeout for inline JSON results. A statement can change data, so it is never resent.
A repeated request may repeat the external action. Review the previous result before retrying a write.
Start a terminated cluster with its last configuration. Nothing happens if it is not terminated.
A repeated request may repeat the external action. Review the previous result before retrying a write.
Use the account and permissions required for your process. Review the connection check before configuring a workflow action.
Select a listed operation and map its required inputs. Use test records and add a human approval before writes that need review.
Inspect step outputs and confirm the result in Databricks. A saved workflow is not a completed provider action.
