Planning a Switchover

You can plan an Application Switchover from the migration planner: the advisory questionnaire that produces a recommended migration plan, not only from the new migration wizard. The planner is advisory: it produces a plan document with a recommendation and a cutover runbook. It does not run anything. Its existing Create Job action then seeds a real switchover job.

Switchover Opt-In

The planner’s Migration Scope question gains an Application Switchover opt-in, a single toggle, mirroring the wizard’s cutover-only opt-in. Off (the default) is a standard migration covering data and/or schemas; on is Application Switchover. The selection persists across steps and when the draft is reopened.

When the toggle is on, the shared producer-visibility note is shown (the same component the wizard and the Begin cutover modal use). It is informational and never blocks proceeding.

Note

Toggling the switchover option off (reverting to a standard migration) restores all the hidden replication steps, brings back the replication cutover controls, and removes the switchover framing and producer note, without stranding you on a step that no longer exists. The switch is reversible at any time on the scope question.

Scope Reframed as Provision Targets

In switchover mode the scope items (Topic Data, Consumer Groups, ACLs) are reframed as provision targets: what the switchover will create empty or seed on the destination, not what will be replicated. The Schemas scope item is hidden: schema provisioning is configured in the wizard, not the planner.

Tip

The topics-required dependency is relaxed in switchover mode: consumer groups and ACLs can be selected independently of topics, since a switchover may legitimately track groups or provision ACLs on a destination whose topics already exist or are out of scope.

Steps Shown and Hidden

In switchover mode the planner shows: Cluster Types, Migration Scope, the source and destination cluster steps, Cutover Planning (reframed), Applications, and Review & Create.

It hides the steps that only make sense for a running replication engine: Tool Preference, Migration Infrastructure, Inter-Cluster Connectivity, Gateway, Gateway Selection, Cost Analysis, AWS Credentials, and Connectors. Two more steps drop out for a different reason: switchover mode turns the Schemas scope item off, and both Schema Registry and Source Serialization are gated on that scope item rather than on the mode itself.

The step sidebar, the progress indicator, and the “Step X of N” counter all reflect only the visible switchover steps: for example “Step 3 of 7”, counting the switchover steps, not the full standard-migration step list.

Note

These are the migration planner’s steps. They differ from the new migration wizard’s steps and hidden-step list because the planner and the wizard are separate surfaces, the planner asks about tooling, cost, connectivity, and gateway that the wizard does not. The difference between the two lists is by design, not an inconsistency.

Cutover Planning Reframed

The Cutover Planning step drops the replication controls (cutover strategy with gateway-assisted, delivery guarantee, downtime tolerance), those assume a replicated backlog a switchover does not have. It asks two things instead, matching the switchover cutover model:

• Cutover order: Coordinated stop (default) or Producers first.
• Cutover granularity: Big bang or Rolling.

The default is Coordinated stop, Rolling. The step explains that producers-first is made safe by draining the source to zero consumer lag before consumers are switched.

The Runbook

Review & Create renders a switchover summary (clusters; topics, ACLs, and consumer groups in scope; the chosen order and granularity) and generates an advisory switchover runbook: provision the destination structure, then the producer prerequisite check, then the cutover steps in the chosen order, then completion. It does not describe a replication engine, lag, or offset translation. Consumer groups are seeded at earliest by default; the summary states the seed policy is adjustable per group in the wizard.

The generated plan’s overview presents the switchover as a migration approach: an informational note that no replication tool is used. The “Recommended Tool” card is absent: because a switchover runs no replication engine, the plan does not present a recommended tool with alternatives the way a data or schema plan does.

Create Job Handoff

The generated plan exposes the existing Create Job action. Clicking it applies the plan into a switchover-mode configuration and opens the new migration wizard pre-filled: the mode is Application Switchover, the clusters and in-scope provision targets are populated, the consumer-group seed policy is set to earliest, and the cutover order and granularity reflect the plan. You can adjust anything before creating the job.

Tip

A switchover job carries no live migration plan. The plan is used once, at Create Job, to seed the configuration, the same one-time seeding that data and schema plans already do. After that the job is a switchover job like any other.

Backward Compatibility

Switchover support in the planner is purely additive: it appears as an opt-in toggle and only changes planner behavior once turned on:

• The standard data migration and schema migration planner flows are unchanged: the same steps, scope behavior, cutover questionnaire (strategy, delivery guarantee, downtime tolerance, gateway), generated plan, and Create Job handoff.
• An existing saved plan that predates switchover opens and behaves exactly as before: nothing about switchover changes a non-switchover plan.