Prepare

Fix what stands between your table selection and a first start: unmarked source tables, missing target tables, and the SQL for a DBA.

The Prepare step is the fourth step of the New pipeline wizard, headed Prepare to run. SQL Flow checks, without changing anything, that the source is ready to be captured and that the target has the tables the pipeline writes to. If something is missing, this step shows the exact SQL and the actions that fix it. If nothing is missing, the wizard moves straight from Tables to Review.

What the check does

The check runs on your agents when you leave the Tables step. It has two parts:

• Source readiness. For SQL Server DirectLog, SQL Server Reconstruction, Oracle DirectLog and Sybase ASE DirectLog, the agent reads the source catalog to check the selected tables. What it looks for depends on the mode. See SQL Server DirectLog, Reconstruction, Oracle DirectLog and Sybase ASE DirectLog.
• Target schema. For relational targets, the agent checks that each selected table exists in the target with a usable key. The same check predicts what a first start does to each table, which the Review step reports.

Kafka, Google Pub/Sub and Fabric Mirror targets, and source monitors, have nothing to prepare, and the step says so.

If the check can't reach a verdict, for example because the agent went offline, an amber message says Couldn't check this pipeline before creating it: <reason> with Check again. This doesn't block Next, and Create checks again.

Fix the source

When the source isn't ready, a card appears with a chip that names the gap:

The Tables tab lists the tables that need attention. The DBA SQL tab shows the statements a DBA runs on the source, and Copy DBA SQL copies them. SQL Flow never runs Oracle statements: an Oracle DBA reviews and runs the SQL in the source PDB.

For SQL Server DirectLog, you can let SQL Flow mark the tables:

1. Read the statements on the DBA SQL tab.
2. Select Mark now (<n>). The dialog Mark these tables for capture? repeats what runs on the SQL Server. If the database had no CDC capture and cleanup jobs, SQL Flow drops the two jobs that marking creates, because DirectLog reads the log directly. Jobs that already exist are left alone.
3. Select Mark tables. The step confirms Marked <n> tables for capture. and checks again.

For Sybase ASE DirectLog, Mark now marks the selected tables for replication with sp_setreptable, using the pipeline login, which needs replication_role. See Sybase ASE DirectLog: Setup.

If the login can't mark tables, the card shows the refusal in red. SQL Flow never changes a server option, so when tables with large-value columns need max text repl size, a red box shows the statement for a DBA, and those tables stay unmarked until the DBA has run it. See Large values and LOBs.

After the DBA acts, or after Mark now, select Check again. The card clears only when the source answers cleanly.

Prepare the target schema

When the target lacks tables, a card marked target schema says Not ready: <n> tables must exist in the <engine> target before this pipeline can run. and lists them. Hover a table to see why it isn't ready. Choose how to fix it:

• Create them yourself. Create the tables and keys in the target, then select Check again.
• Let SQL Flow create them. Select Show SQL to read exactly what will run, and Copy SQL to copy it. Then select Prepare target schema and confirm with Prepare schema. The step reports Target schema prepared: <n> table(s) created in the target. and checks again.

This action creates tables and their keys, and nothing else. It doesn't add a column to a table that exists, so fix a table that lacks a column by hand. It also doesn't create the pipeline: you do that on the Review step.

Leave the step

Next is enabled when the check has answered for the current choices and no source or target gap remains. Otherwise the red box says what to do, for example Have a DBA run the SQL below in the source, then check again. or Prepare the target schema below, or create the tables in the target yourself and check again.

If Create later finds a gap on the Review step, the wizard returns you to this step.

See also

• Tables and configure
• Review and create
• Large values and LOBs
• Encrypted data support