A Deep Dive Into Ingesting Debezium Events From Kafka With Flink SQL

A detailed comparison of the different options for ingesting Debezium CDC events from Kafka into Flink SQL. Covers three main approaches: the Apache Kafka SQL Connector in append-only mode (treating events as a raw stream), the same connector in changelog mode using the debezium-json format (for stateful streaming queries and materialized views), and the Upsert Kafka SQL Connector (for flat events after applying the ExtractNewRecordState SMT). Each approach is explained with schema configuration examples, operational semantics, trade-offs around metadata preservation, write amplification, and appropriate use cases. Also touches on Flink CDC as an alternative to Kafka-based ingestion.