Zero-downtime migration

Move the starting state. Then close the gap while production stays open.

SQL Flow pins the change position before the initial load, copies tables in parallel, and keeps committed changes in a durable stream until the target catches up.

Talk through my use case See Seed + Stream

One coordinated path

Initial load and CDC share the same migration boundary.

The pipeline records where continuous capture begins, copies the selected tables, and applies every committed change that arrived during the copy.

  • ✓Per-table progress, rate, and durable completion markers
  • ✓Parallel table slices with source-aware snapshot behavior
  • ✓Catch-up from the pinned source position
  • ✓Explicit reload and reseed when the plan changes

Cut over with evidence

Know the remaining lag before you move traffic.

Separate capture and apply lag make the cutover decision observable. Let the target reach the durable frontier, validate it, then switch the application against a checkpoint you can name.

  • ✓Visible source and destination positions
  • ✓Target preflight before the first load
  • ✓Independent cutover for multiple destinations

The payoff

See what the capability changes.

Handoff

Pin before copying

Establish the CDC handoff before the first table slice moves.

Parallel

Load in parallel

Use bounded workers and range slicing to seed large targets.

Durable

Catch up durably

Replay every concurrent change before moving the application.

Take the next step

Turn cutover into a measured handoff.

Bring the source, target, table sizes, and outage budget. We will map the load, CDC handoff, validation, and cutover sequence.

Talk through my use case