This library provides an implementation of the Transactional Outbox Pattern based on polling events from a relational database and publishing them to a message broker. On the receiving side, an inbox implements the Idempotent Consumer Pattern: it records the ids of consumed events, so a redelivered event is skipped.
The library offers flexible configuration for processing different event types. Each event type is handled independently, and events are processed in parallel using a configurable thread pool, allowing predictable throughput and controlled resource usage.
This approach ensures reliable event publication without relying on database log-based CDC solutions, making it suitable for environments where simplicity, portability, and explicit control over event processing are preferred.
The full documentation lives here:
- Quickstart - add the dependencies, configure an event type, publish and consume it.
- Delivery Guarantees - what is guaranteed, in which failure, and what is not.
- Configuration - every property with its default.
- Schema Migrations - creating the tables with Flyway or Liquibase.
- Atomic persistence - stores events within the same business transaction, ensuring consistency.
- Adaptive polling - dynamically adjusts the polling interval based on workload to optimize database usage.
- Flexible polling configuration - allows customization of polling intervals, retry policies, and backoff strategies per event type.
- At-least-once - guarantees outbox delivery with configurable retries and backoff policies.
- Effectively-once - optionally records an incoming event in the inbox within the same transaction as its handling, so a redelivered event never changes data twice.
- At-most-once - optionally records an incoming event in the inbox before handling it, so a side effect such as sending an email is never repeated.
- Dead Letter Queue - stores events that fail after all retry attempts, with REST API support for management.
- Background service - handles recovery of stuck events and cleanup of processed or consumed events automatically.
- Observability - provides out-of-the-box metrics integration via Micrometer.
- Schema migrations - ships its tables as versioned migrations, applied with Flyway or Liquibase or plain SQL scripts.
- Databases: PostgreSQL 16+, MySQL 8.0.13+, Oracle 23+.
- Schema Migrations: Flyway, Liquibase.
- Message Brokers: Apache Kafka 3.7+, RabbitMQ 3.12+.
- Cache Storage (optional for caching): Redis 7+.