Sources and destinations

A source is a Connection with the SOURCE capability, used for database discovery, readiness and capture. A pipeline selects tables and owns a Debezium source connector. A destination is a Connection with the DESTINATION capability. A delivery links pipeline topics to that target and owns an independent Kafka Connect sink connector.

One capture pipeline can fan out to several deliveries. Pausing a delivery does not pause capture. Each delivery has independent consumer offsets, so a new delivery can replay retained Kafka history.

PostgreSQL/MySQL deliveries use JDBC table mappings. AWS S3/MinIO deliveries archive CDC envelopes in JSONL files; they have topic selection, compression and file settings, with no SQL table or primary-key options. Each delivery has independent connector lifecycle and offsets.