Pipeline wizard
Review and create
Name the pipeline, review what it will do to the target, and create it paused or create and start it in one step.
The Review step is the last step of the New pipeline wizard, headed Review & name. You name the pipeline, check a summary of your choices, see what a start will do to rows already in the target, and create the pipeline. You can create it paused, or, when every readiness check is green, create it and start it in one action.
Name the pipeline
Enter a name in Pipeline name. The wizard suggests one from the source and target databases, for example orders → analytics.
- The name must contain at least one letter or digit.
- It can be up to 100 characters, counting each run of spaces or symbols as one.
- Names are unique in the account. If one exists, the field says
A pipeline named '<name>' already exists.
Check the summary
The summary card shows the two engines, such as SQL Server → PostgreSQL. Select Show details to expand it.
| Row | What it shows |
|---|---|
| Source, Target, Selected tables | Your choices, each with an Edit link back to its step. |
| Source encryption, Target encryption | Encryption: Require or Encryption: Prefer (in amber), for connections that have the choice. |
| Capture method | The capture mode, such as DirectLog. |
| Source access | verified through <agent> after a passing test, the source test failed: fix it, or the pipeline will not run, or not tested in this wizard. |
| Initial load | What a start copies. For Kafka, whether existing rows are published first or only changes are streamed. With Stream changes only, Source scan adds that the initial load still reads every tracked table. |
| Target handled by | Which agent holds the target credentials and writes to it. |
| Capture agent, Apply agent, Relay port, Relay trust, Relay dial | Shown when you chose a relay: the two agents, the port, TLS 1.3, mutual, and verified or the reason the dial isn't verified. |
| Apply | Every row version, in order, or Intermediate row versions collapsed, if preflight proves the target safe. See Collapse row versions. |
A Configure card lists the table names, column names, row filters and output columns you set. Its Configure link returns to the Tables step with the dialog open.
See what a start does to existing rows
When the Prepare check could reach the target, a card summarizes what a start does to each table. Its headline is amber when rows would be replaced: Starting this pipeline will truncate <n> tables on the target. Expand the card to see the tables in groups:
| Group | What happens |
|---|---|
| Truncated, then loaded again | The table has rows. A start empties it and copies the source table. |
| Created, then loaded | The table doesn't exist yet. Create it with Prepare target schema on the Prepare step before the first start. |
| Empty, loaded | The table is empty. It is loaded. |
| Could not be checked | The check couldn't answer for this table. |
SQL Flow won't empty a SQL Server, Azure SQL or Fabric SQL target table whose rows it can't prove are its own, for example when the source is PostgreSQL, MySQL, Oracle or Sybase. The card says so: empty such a table yourself before the start. See Rows already in the target.
The card shows the target as it is now. Creating the pipeline paused doesn't change those rows. The first start does, whether you start it here or later on the pipeline's page.
Read the notices
An amber or red notice at the bottom lists what the wizard already knows, such as a failed source test, an offline agent, a connection not tested in this wizard, or an unassigned pipeline. The notices don't block creating the pipeline.
Create the pipeline
Create and start is offered as the primary button when the pipeline can run the moment it exists:
- an agent is assigned and online, and for a relay both agents are online;
- the source and the target tested OK in this wizard (a source monitor needs only the source);
- the target isn't missing tables, and the source isn't waiting for a fix.
Otherwise Create paused is the only button.
Create paused creates the pipeline in the Paused state. Nothing is loaded or captured until you start it from the pipeline's page. Create and start creates the pipeline and starts it.
When you select either button:
- For Create and start, if target tables hold rows that a start replaces, the dialog Replace existing target rows? names them and the target. Select Replace rows and start to continue, or Keep editing to go back. Tables SQL Flow won't empty are marked as left out: empty them yourself before the start.
- The wizard checks source and target readiness again, and returns you to Prepare if it finds a gap. If your source test is a few minutes old, it tests the source again, and stops with the test's message if that fails.
- The pipeline is created and, for Create and start, started.
You then land on the pipeline's page. If the start wasn't confirmed, the page shows Pipeline created, but Start was not confirmed. with the reason. Check the pipeline's status before you select Start again. If the create fails, the red box shows the reason and nothing is created.
Verify
On the pipeline's page, a started pipeline shows Loading with a progress bar for the initial load, then Healthy when it streams. A paused pipeline shows Paused and a Start button. See Pipeline detail and Lifecycle.
