This commit introduces a comprehensive refactoring of the messaging module to establish a fully reactive, non-blocking, and robust infrastructure for Kafka-based communication.
Features & Refinements
Reactive Publisher:
The KafkaEventPublisher has been refactored from a blocking implementation (KafkaTemplate) to a fully non-blocking, reactive one using Spring's ReactiveKafkaProducerTemplate.
The EventPublisher interface now returns reactive types (Mono, Flux) to reflect the asynchronous nature of the operations.
Reactive Consumer:
A new KafkaEventConsumer has been implemented, providing a standardized, reusable, and reactive way for services to consume events.
It encapsulates the complexity of reactor-kafka and exposes a simple receiveEvents<T>(topic) method that returns a Flux<T>.
Architectural Cleanup:
The Spring configuration has been split. The basic ProducerFactory and consumer properties reside in messaging-config, while the reactive-specific ReactiveKafkaProducerTemplate bean is now correctly located in messaging-client.
Testing
Added Kafka Integration Test: A new KafkaIntegrationTest has been created to ensure the reliability of the messaging infrastructure.
The test uses Testcontainers to spin up a real Apache Kafka broker for end-to-end validation.
Project Reactor's StepVerifier is used to test the reactive streams deterministically, avoiding flaky tests.
The test correctly manages the lifecycle of Kafka producers to ensure clean shutdown without hanging threads.
Bug Fixes
Resolved UninitializedPropertyAccessException in tests by making the KafkaConfig test-friendly.
Fixed IllegalStateException related to Testcontainers lifecycle by making the container a static resource.
Corrected compilation errors in tests related to resource cleanup by using the concrete DefaultKafkaProducerFactory type.
This commit introduces a comprehensive refactoring of the event-store module to guarantee data consistency and improve the quality and reliability of its test suite.
Data Consistency
Implemented Redis transactions (MULTI/EXEC) for the appendToStream operation in RedisEventStore.
This ensures that writing an event to the aggregate-specific stream and the global "all-events" stream is an atomic operation, preventing data inconsistencies in case of partial failures.
Improved error handling by invalidating the local stream version cache on transactional failures.
Testing Enhancements
Refactored Consumer Tests: Replaced the asynchronous, thread-based consumer test in RedisIntegrationTest with a synchronous, deterministic approach. The test now manually calls pollEvents() to verify event consumption, making it faster and 100% reliable by removing Thread.sleep and CountDownLatch.
Simplified Test Events: Reduced boilerplate code in test event data classes (TestCreatedEvent, TestUpdatedEvent) in both RedisEventStoreTest and RedisIntegrationTest by using the @Transient annotation on overridden properties from BaseDomainEvent.
Fixed Compilation Errors: Resolved various compilation errors in the test suite that arose from refactoring and incorrect mock definitions.
stefan
stefan