Skip to main content

Forcing a Re-Baseline

A re-baseline reloads all rows from the source for a table, replacing the target's current state.

When to re-baseline​

  • After a suspected data drift
  • After a schema change that a target can't handle incrementally
  • After restoring a database backup
  • After dropping and recreating a replication slot

Trigger via CLI​

# Re-baseline a specific table
laredo reload public.config_document

# Re-baseline all tables on a source
laredo reload --source pg_main --all

# Re-baseline everything
laredo reload --all

What happens during re-baseline​

  1. Engine pauses the change stream (stops reading, keeps connection open)
  2. Engine calls OnTruncate on all targets for the table (clears existing state)
  3. Opens a new REPEATABLE READ snapshot transaction
  4. Executes SELECT * for each table, delivers rows via OnBaselineRow
  5. Calls OnBaselineComplete
  6. Resumes the change stream
  7. Applies any changes that accumulated during the reload
  8. ACKs the current position once reload and catch-up are complete

Some changes may overlap with rows just loaded during the reload window. Targets must be idempotent for this brief overlap.

Resetting a source​

For a full reset (drops and recreates the replication slot):

laredo reset-source pg_main

This is more drastic than a reload — it destroys all position tracking. Use it when the slot is corrupted or you're changing the publication configuration.

To also drop the publication (for decommissioning):

laredo reset-source pg_main --drop-publication