Change Data Capture (CDC)

Row triggers fire inside the application's own transaction and write an extra audit row per change, so every write becomes two. Commit latency, lock duration, log volume and failure surface all grow, and the audit table needs its own drain and purge.

Expand it when consumers need a real schema for the event body — typically with a registry-backed converter. Leave it as a string when the payload's shape varies between events, because the structure is inferred per record and inconsistent shapes produce unstable schemas.

It replaces the default SELECT that reads a table during the snapshot with a statement you supply, so you can filter rows or restrict columns. It affects the snapshot only; streaming afterwards still captures every change to that table.

The row's history is keyed by the old value and its new event by the new one, so the sink upserts a fresh row and leaves the old one behind as an orphan. Key on an immutable identifier instead.

The connector writes a marker into the log, reads one key-range chunk, writes a second marker, then drops from the chunk any key that changed between the markers — the live stream already carries the truth for those. Streaming never pauses.

Debezium Server is a standalone process that streams change events to non-Kafka sinks; the embedded engine is a library running inside your own application. Choose them when there is no Kafka Connect cluster, accepting that you own restarts, scaling and offset storage.