A production-style reference implementation of a Java/Spring Boot microservices system, modeled on patterns used in real banking and wealth-management platforms (client onboarding, document workflows, approval flows, and secure API access).
Most public Spring Boot demos stop at a single CRUD service. This repo is meant to show how those services actually fit together in a regulated, enterprise environment — service boundaries, async messaging, security, and observability — the way they're built in banking systems rather than in tutorials.
┌─────────────┐ ┌──────────────────┐ ┌─────────────────┐
│ API Gateway │────▶│ Auth Service │────▶│ User/Profile │
│ (Spring │ │ (JWT/OAuth2) │ │ Service │
│ Cloud GW) │ └──────────────────┘ └─────────────────┘
└──────┬───────┘
│
▼
┌─────────────────┐ ┌──────────────────┐
│ Onboarding │────▶│ Kafka │
│ Service │ │ (event bus) │
└─────────────────┘ └────────┬─────────┘
▼
┌──────────────────┐
│ Document / │
│ Approval Service │
└──────────────────┘
Design decisions worth noting:
- Kafka over synchronous REST for onboarding events — decouples the onboarding, document, and approval services so a slow downstream consumer (e.g. document generation) never blocks the onboarding flow. Chosen over RabbitMQ for built-in event replay/audit trail, which matters for regulatory traceability.
- JWT/OAuth2 at the gateway, not per-service — centralizes token validation, reduces duplicate auth logic across services, and simplifies key rotation.
- Database-per-service — each service owns its schema (Oracle/PostgreSQL) to avoid tight coupling; cross-service data needs are handled via events, not shared tables.
| Layer | Technology |
|---|---|
| Backend | Java 17, Spring Boot 3, Spring Cloud Gateway |
| Messaging | Apache Kafka |
| Security | Spring Security, JWT, OAuth2 |
| Persistence | Spring Data JPA/Hibernate, PostgreSQL/Oracle |
| Testing | JUnit 5, Mockito, Testcontainers |
| API Docs | OpenAPI/Swagger |
| Build/CI | Maven, GitHub Actions |
- At small scale, database-per-service adds operational overhead (multiple schemas to manage). At larger scale it pays off by letting services deploy independently — the right call once you have more than ~5 services.
- Synchronous REST between gateway and auth service is a single point of latency; a cached token-validation layer (e.g. local JWKS caching) would reduce gateway load in high-throughput scenarios.
docker-compose up -d # starts Kafka, Postgres, Zookeeper
mvn clean install
mvn spring-boot:run -pl auth-service
mvn spring-boot:run -pl onboarding-serviceImplemented:
- JWT/OAuth2 authentication flow
- Kafka producer/consumer for onboarding events
- REST APIs for onboarding, document, and approval services
- Unit and integration tests (Testcontainers for Kafka/Postgres)
Planned:
- Distributed tracing (OpenTelemetry)
- Circuit breaker pattern (Resilience4j) for downstream service calls
- Kubernetes deployment manifests
Built by Santhosh Kumar K — Senior Java Backend Engineer / Technical Lead, 9+ years in banking & financial services (CIBC, Standard Chartered, Credit One Bank).