Skip to main content

Build your first pipeline

Where: sidebar → Jobs, then open a pipeline in its Data Catalog database.

This walkthrough uses illustrative names: source training-orders-source, target database training, and target table orders_copy. These resources are not installed by the documentation. Use an approved small source table and an isolated target in your own workspace.

Before you start

Create the job

Open Jobs and select New pipeline. Choose the Data Catalog database, enter a Pipeline name, and submit the dialog. Open the created pipeline.

Configure a source node

In the visual editor's Graph nodes catalogue, choose a source node supported by your connector. Select the verified connection and source relation in its property panel. Use the fields offered by that node; do not paste a database password into a SQL transformation.

Configure the target

Add a supported target node and connect the source output to it. Set the target database and table. Review the target write mode and any schema-related settings before saving.

For the first exercise, use a new training table rather than an existing production table. Transformations can be added after the simple source-to-target path works.

Save and review

Save the graph and resolve validation errors. Inspect node properties and the generated script where available. The generated script represents a saved version, not necessarily unsaved canvas edits.

A node preview requires a saved pipeline and is a limited preview, not proof of a completed target write.

Run the saved pipeline

Select Run when available. Note the execution identifier and open Job running. Wait for Succeeded; Queued, Submitted, and Running are intermediate states.

Verify the managed table

Open the target database under Data Catalog, locate the table, and inspect its columns and preview. Run a small query if authorized. Compare a few known source records and confirm that the write mode produced the intended result.

Safe iteration

Re-running a pipeline can append duplicates or replace data, depending on the target configuration. Confirm the write mode and expected output before every retry. Cancellation does not guarantee that already-written data is rolled back.

Use Clone to explore changes in a separate pipeline. Archive changes pipeline availability; it is not a command to drop its target tables.

If another editor saved a newer version, review or reload that version before continuing. Do not repeatedly overwrite a conflict without understanding the other changes.

Continue with Monitor runs or Query your data.