Open-source Oracle GoldenGate alternatives

Debezium is an open-source platform for change data capture. You configure it to monitor your databases, and your applications then consume an event for each row-level change. Only committed changes are visible, so consumers do not need to deal with transactions that are rolled back.

It provides one consistent model for change events across different database systems, so applications are insulated from the details of each engine. Changes are recorded in durable, replicated logs, which means a consumer can be stopped and restarted and still receive every event it missed. This removes the need for database-specific triggers or ad hoc change-monitoring code, which the project describes as hard to get right when ordering and low impact on the database both matter.

Debezium is built with Java and Kafka Connect in mind, as the project topics show, and is licensed under Apache-2.0, with the ANTLR grammars in its DDL parser module under the MIT licence. It runs as part of your own data infrastructure and suits data engineers building event streaming pipelines, cache invalidation or data replication.

Dozer is a real-time data movement platform that uses change data capture (CDC) to stream data continuously from source databases into destinations such as data warehouses. It is aimed at data engineering teams who need faster, simpler CDC pipelines than assembling Debezium and Kafka themselves, with native support for stateless transformations along the way.

It supports a range of source connectors including Postgres, MySQL, Snowflake, Kafka, MongoDB, Amazon S3, Google Cloud Storage, Oracle and Aerospike, with extraction and resuming support varying by connector and some sources reserved for the enterprise edition. On the sink side, it supports ClickHouse, Postgres, MySQL and BigQuery in the open-source edition, with Oracle and Aerospike sinks reserved for enterprise. The project notes its own typical use is moving data into ClickHouse to build data APIs and integrations with LLMs.

Dozer is configured through a single configuration file and is written in Rust for performance, claiming to be significantly faster than a Debezium-plus-Kafka setup for comparable CDC workloads. It is released under the AGPL-3.0 license, with full documentation available for setup and configuration details.