Replication

In this article I want to share problems I faced while running Debezium in production with PostgreSQL, and solutions where they exist. This material is based on a talk I gave at SmartData 2025. I assume that you already know what CDC is and how Debezium works in general — if not, start with PostgreSQL: Log-based CDC using debezium and Kafka-connect: overview. All examples below were tested with PostgreSQL 15, 16 and 17. If a version is not mentioned explicitly, PostgreSQL 17 is assumed. Examples are implemented using a runtime wrapper around Debezium Engine, but the same problems and solutions apply to the Kafka Connect connector, Debezium Server, and other deployment options. ...

In this little article I’ll show different ways to set up debezium for log-based CDC. Before diving into details about debezium, I’ll shortly describe CDC and why it may be helpful in some tasks. CDC: Change Data Capture In the Internet the CDC is described as a design pattern which allows to track data changes (deltas). Let’s consider this approach on table user_balances. Initial state of table is: ...

Prerequisite All examples assume that postgresql is already installed on your machine. Also, all examples are created using PostgreSQL 14.1 on aarch64-apple-darwin20.6.0, compiled by Apple clang version 13.0.0 (clang-1300.0.29.3), 64-bit. Log shipping replication Log shipping replication (i will use a short name for it LSR) is another one method to physically replicate data between multiple database clusters. As name says this method is about to replicate data through WAL-files (segment) which is transferred between instances. This is probably the most simple and straightforward method for data replication, but this simplicity comes with price and compromises which also should be accounted. ...

Prerequisite All examples assume that postgresql is already installed on your machine. Also, all examples are created using PostgreSQL 14.1 on aarch64-apple-darwin20.6.0, compiled by Apple clang version 13.0.0 (clang-1300.0.29.3), 64-bit. Streaming replication Streaming replication is a built-in mechanism in PostgreSQL to replicate data between multiple servers. It is a low-level replication mechanism as it streams WAL data from primary server to the replica through the physical replication slot, so it is highly recommended to replicate data between servers using similar PostgreSQL major version (minor versions could be different). Also, it is a good idea to have equal servers in terms of server configuration such as CPU, RAM and Disks, especially if you consider to promote replica to master if primary server goes down. ...

Prerequisite All examples assume that postgresql is already installed on your machine. Also, all examples are created using PostgreSQL 14.1 on aarch64-apple-darwin20.6.0, compiled by Apple clang version 13.0.0 (clang-1300.0.29.3), 64-bit. Logical replication Logical replication is another method to replicate data between multiple nodes. This replication uses publish-subscribe model. Each publisher may have multiple subscribers and each subscriber can subscribe to multiple publisher. Also, each subscriber may be a publisher for another node which make it possible to create a cascading replication. ...