Cheat Sheets

REST API status codes

// Problem+JSON error body (RFC 7807)
{
  "type": "https://api.example.com/errors/insufficient-stock",
  "title": "Insufficient stock",
  "status": 409,
  "detail": "SKU-42 has 0 units available",
  "instance": "/orders/req-8f2a"
}

Spring Cloud annotations

@FeignClient(name = "inventory-service", configuration = FeignConfig.class)
public interface InventoryClient {
  @GetMapping("/api/v1/stock/{sku}")
  StockResponse getStock(@PathVariable String sku);
}

@Bean
@LoadBalanced
RestTemplate restTemplate() { return new RestTemplate(); }

Resilience4j annotations

// Decorator order (outer → inner): RateLimiter → CircuitBreaker → Bulkhead → Retry → TimeLimiter → call

@CircuitBreaker(name = "payment", fallbackMethod = "payFallback")
@Retry(name = "payment")
@Bulkhead(name = "payment", type = Bulkhead.Type.SEMAPHORE)
public PaymentResult charge(Order order) { ... }

// application.yml
resilience4j.circuitbreaker:
  instances:
    payment:
      slidingWindowSize: 20
      failureRateThreshold: 50
      waitDurationInOpenState: 30s

Kafka consumer config

@KafkaListener(topics = "order-placed", groupId = "inventory-service")
public void onOrderPlaced(ConsumerRecord<String, OrderPlacedEvent> record, Acknowledgment ack) {
  process(record.value());
  ack.acknowledge();  // after idempotent success
}

Kafka producer config

spring:
  kafka:
    producer:
      acks: all
      properties:
        enable.idempotence: true
        max.in.flight.requests.per.connection: 5
    template:
      default-topic: domain-events

Pattern decision tree

flowchart TD
  START[New integration problem] --> Q1{Migrating monolith?}
  Q1 -->|yes| STR[Strangler Fig plus ACL]
  Q1 -->|no| Q2{Multi-service write?}
  Q2 -->|yes| SAG[Saga plus Outbox]
  Q2 -->|no| Q3{Need instant read-your-writes?}
  Q3 -->|yes| SYNC[Sync REST or gRPC plus Timeout Breaker]
  Q3 -->|no| ASYNC[Events Kafka plus Idempotent consumer]
  SYNC --> Q4{Client-specific API?}
  Q4 -->|yes| BFF[BFF or API Composition]
  Q4 -->|no| DONE[Ship with correlation ID and RED metrics]

Sync vs async guide

Rule of thumb: if the user did not trigger it and does not wait — async.
If failure blocks the UI — sync with resilience, not fire-and-forget.

CAP theorem cheat

PACELC: If Partition → A or C; Else → Latency or Consistency.
Document per-service in ADR — not one slide for whole platform.

Data pattern selection guide

Combine: DB per service + outbox + saga + idempotent consumers
is the default production stack for write paths.

Saga vs 2PC decision

Orchestrated saga: central coordinator — easier trace, coupling risk.
Choreographed saga: domain events — decoupled, harder debug.
Both need: outbox, idempotent consumers, explicit compensations.

Service mesh comparison

Observability stack options

Minimum viable: Micrometer + Prometheus + Grafana + OTel traces + JSON logs + Loki.
Gate prod on: RED dashboards, one SLO, trace_id in every log line.

Deployment strategy comparison

Never kubectl set image in prod without Git record.
Canary promotion gates: error rate, p99 latency, business metric (conversion).