Schema Provisioning

An Application Switchover job can optionally provision schemas onto the destination Schema Registry as part of the same one-shot provision that creates topics, ACLs, and consumer-group offsets. Production teams often disable producer schema auto-registration, so the destination must already hold the schemas or apps fail at cutover. Schema provisioning closes that gap. It is opt-in and off by default, and it never copies message data.

Opt-In Step

The switchover wizard always shows a Schema Provisioning step with a single enable toggle. With the toggle off, the step shows only that control — no connection capture, no subject selection, no Schemas monitor tab, and no schema results in the report. The job behaves exactly as described in Provisioning the Destination.

With the toggle on, the wizard shows source and destination Schema Registry connection capture (with the existing connection check) and a subject selection step.

Capability Gate

When you enable schema provisioning and run the connection check, the product detects whether the source registry supports exporters by non-invasively listing its exporters with the supplied credentials. The check never creates or deletes an exporter.

Source registry	Result
Confluent Platform 7.0+ or Confluent Cloud (returns an exporter list)	Capable — schema provisioning is available.
No exporter API (e.g. Confluent Platform older than 7.0), or access denied	Not capable — schema provisioning is blocked here.

Caution

If the source is not exporter-capable, schema provisioning is blocked at connection time with a clear message to pre-stage schemas on the destination manually — never as an opaque mid-provision failure. The rest of the switchover is unaffected and the job can still be created without schema provisioning.

The destination registry must also be reachable and writable; this is part of the same check.

Subject Selection

Subject selection supports an explicit list, wildcard patterns, and exclusion patterns. Each subject row shows a destination badge: will create when absent on the destination, or exists on destination when already present.

What Provisioning Copies

With schema provisioning enabled, the one-shot provision copies the selected subjects to the destination, carrying each subject type (Avro, Protobuf, or JSON Schema), all versions, references, and the compatibility level. References are created before the subjects that depend on them.

Caution

The destination assigns its own schema IDs — source IDs are not preserved. This is correct for a switchover, because no source data lands on the destination. If you plan to reload source-produced data that carries embedded source schema IDs, those IDs will not match on the destination; ID-preserving import is not available in this version.

No schema exporter or background process remains running after provision completes. Compatibility is set only at creation, never changed on an existing subject; a subject with no explicit source setting takes the destination default.

Conflicts (Brownfield)

A subject already on the destination is skipped, never overwritten. A subject present with a schema incompatible with the source is reported as a conflict with a readable reason so you can resolve it on the destination.

Tip

A flagged conflict does not block cutover — it stays visible in the report. Resolve it on the destination registry when you are ready.

Report and Monitor

The provision report includes schema results across four outcomes: created, skipped, failed, and conflict, with schema totals in the summary. A Schemas monitor sub-tab appears (only when schema provisioning is enabled) listing each subject with its result and a readable reason on non-created rows. The tab is static — a schema does not move at runtime, so there is no live cutover column. Failed subjects are selectable for selective reprovision (with a Retry all failed shortcut); a conflict is not retryable.

Sync Now

Sync now creates subjects newly in scope (for example a new source subject matching a wildcard) and skips those already created. Schema sync is grow-only — it never deletes a destination subject.