Home → NOTES → Event-Driven Architecture

Event-Driven Architecture

category: Architecture
tags: event-driven, kafka, java, kubernetes, microservices

Main Topic 1

Event-Driven Architecture (EDA) is a design paradigm where components communicate through events—emitted, consumed, and reacted upon—decoupling the services.

Subtopic A: What It Is

• Asynchronous communication – Producers emit events, consumers listen and act
• Loose coupling – Services do not directly call each other
• High scalability – Suits high-throughput and reactive systems

Subtopic B: Why Use It

• Enables real-time processing and responsiveness
• Decouples services for better fault tolerance
• Facilitates audit trails and logging through events

Subtopic C: When To Use It

• Real-time systems (e.g. fraud detection, analytics)
• Microservices needing loose coupling
• Need for async workflows and horizontal scaling

Main Topic 2

Code Example (Java + Spring Kafka)

@Component
public class OrderListener {
    @KafkaListener(topics = "orders")
    public void handleOrder(String orderJson) {
        // process order
    }
}

Infra Example (Kubernetes + Kafka)

apiVersion: apps/v1
kind: Deployment
metadata:
  name: kafka-consumer
spec:
  replicas: 2
  selector:
    matchLabels:
      app: kafka-consumer
  template:
    metadata:
      labels:
        app: kafka-consumer
    spec:
      containers:
      - name: consumer
        image: myregistry/kafka-consumer:latest
        env:
        - name: KAFKA_BOOTSTRAP_SERVERS
          value: kafka:9092

Key Concepts Summary

• Events decouple producers and consumers
• Supports async flows, retry patterns
• Event brokers like Kafka or RabbitMQ act as middleware

Best Practices / Tips

1. Design events as immutable facts
2. Use schema validation (e.g., Avro)
3. Handle idempotency in consumers

Common Issues / Troubleshooting

Problem 1: Event loss or duplication

• Cause: Consumer restarts or network partition
• Solution: Use offset management and idempotent processing

Problem 2: Hard to trace event flow

• Cause: No centralized logging or monitoring
• Solution: Use distributed tracing and observability tools

References / Further Reading

• https://martinfowler.com/articles/201701-event-driven.html
• https://kafka.apache.org/documentation/

Last updated: 2025-08-26 20:00 UTC