Notification System - for learning

This project is sample of a Notification System built with Spring Boot (Java) and RabbitMQ, containing:

• User notifications (in-app, push mock)
• Broadcast messages
• News notifications with hash-based partitioning (preserves ordering per category)
• Dynamic listener registration (method 2) — one container per partition, registered programmatically
• Dead Letter Queues (DLQ) per partition
• Resilience4j: Circuit Breaker & Rate Limiter
• Log tracing entity (LogTrace) saved to DB
• SLF4J + custom logback-spring.xml using MDC keys: userId, notificationId
• H2 in-memory DB for quick testing
• Docker Compose to run RabbitMQ locally

Structure

See src/main/java/... for code. Key packages:

• config : RabbitMQ configuration and partition setup
• entity : Notification, UserPreference, LogTrace
• repository : Spring Data JPA repositories
• service : Publisher, Dynamic listener registrar, Notification service, Worker, Push sender
• controller : REST endpoints to send notifications and publish news

Running locally

1. Start RabbitMQ with Docker Compose: docker-compose up -d
2. Build and run the Spring Boot app: mvn spring-boot:run
3. Open RabbitMQ management UI: http://localhost:15672 (guest/guest)
4. Use these endpoints:
   • POST /api/notifications/send?userId=1&title=Hello&message=Hi
   • POST /api/notifications/broadcast (body: JSON array of userIds, params title & message)
   • POST /api/news/publish?categoryId=2&title=News&content=Hello
   • GET /api/notifications/user/{userId} to query recent messages

Important implementation notes

• Partitioning / Ordering: Publisher calculates a partition based on categoryId % NUM_PARTITIONS and publishes to news.partition.{i} routing key.
• Dynamic listeners: DynamicNewsListenerRegistrar registers a SimpleMessageListenerContainer for each partition. Each container uses 1 consumer to preserve ordering inside that partition queue, while having multiple partitions allows parallelism.
• DLQ: Each partition queue is created with x-dead-letter-routing-key set to news.partition.{i}.dlq and those DLQ queues are also declared.
• Resilience4j: NotificationService uses @CircuitBreaker and @RateLimiter to protect DB and enqueue operations.
• Logging: Use MDC to inject userId and notificationId into logs. logback-spring.xml prints them.
• Trace logs: All important actions are saved to LogTrace table for audit.

Links / References

• Designing a Notification System (DesignGurus): https://www.designgurus.io/course-play/grokking-system-design-interview-ii/doc/designing-a-notification-system , https://medium.com/@tanushree2102/designing-a-scalable-notification-system-from-hld-to-lld-e2ed4b3fb348
• RabbitMQ message ordering (Medium): https://medium.com/batc/rabbitmq-message-ordering-on-multiple-consumers-6f484858f589