Monitor Spring Boot API Latency with Actuator, AOP, and Prometheus

1. Introduction

This article explains how to monitor arbitrary API call latency in a Spring Boot application by combining Spring Boot Actuator, a custom @Monitor annotation, AOP, and Prometheus.

2. Environment Setup

Required dependencies:

<dependency>
  <groupId>org.springframework.boot</groupId>
  <artifactId>spring-boot-starter-actuator</artifactId>
</dependency>
<dependency>
  <groupId>io.micrometer</groupId>
  <artifactId>micrometer-registry-prometheus</artifactId>
  <scope>runtime</scope>
</dependency>

Expose all actuator endpoints:

management:
  endpoints:
    web:
      base-path: /ac
      exposure:
        include: '*'

3. Practical Example

3.1 Define the @Monitor annotation

@Retention(RetentionPolicy.RUNTIME)
@Target(ElementType.METHOD)
public @interface Monitor {
  String[] tags() default {};
}

3.2 Implement the monitoring aspect

@Component
@Aspect
public class MonitorAspect {
  private final MeterRegistry meterRegistry;
  private static final String API_TIMER_METER_NAME = "myapp.api.timer";

  public MonitorAspect(MeterRegistry meterRegistry) {
    this.meterRegistry = meterRegistry;
  }

  @Pointcut("@annotation(monitor)")
  private void pcMonitor(Monitor monitor) {}

  @Around("pcMonitor(monitor)")
  public Object around(ProceedingJoinPoint pjp, Monitor monitor) throws Throwable {
    Timer.Sample sample = Timer.start(this.meterRegistry);
    String[] tags = monitor.tags();
    Object ret = null;
    Throwable ex = null;
    try {
      ret = pjp.proceed();
    } catch (Throwable th) {
      ex = th;
      throw th;
    } finally {
      List<String> listTags = new ArrayList<>();
      listTags.addAll(Arrays.asList(tags));
      if (Objects.nonNull(ex)) {
        listTags.add(ex.getClass().getSimpleName());
      }
      Timer timer = meterRegistry.timer(API_TIMER_METER_NAME, listTags.toArray(new String[0]));
      sample.stop(timer);
    }
    return ret;
  }
}

3.3 Service layer

@Service
public class UserService {
  private static final List<User> DATAS = List.of(
      new User(1L, "张三", "男", 22),
      new User(2L, "李四", "男", 23),
      new User(3L, "王五", "女", 22),
      new User(4L, "赵六", "男", 32));

  public List<User> queryUsers() {
    sleep(2000);
    return DATAS;
  }

  public User queryById(Long id) {
    sleep(1000);
    return DATAS.stream().filter(user -> user.getId() == id).findFirst().orElse(null);
  }

  private void sleep(int time) {
    try {
      TimeUnit.MILLISECONDS.sleep(new Random().nextInt(time));
    } catch (InterruptedException e) {}
  }
}

3.4 Controller layer

@Monitor(tags = {"UserController", "list"})
@GetMapping("")
public List<User> list() {
  return this.userService.queryUsers();
}

@Monitor(tags = {"UserController", "ById"})
@GetMapping("/{id}")
public User queryById(@PathVariable Long id) {
  return this.userService.queryById(id);
}

Note: The tags attribute must contain an even number of strings because they are paired as key/value tags.

4. Testing and Visualization

Invoke the two endpoints repeatedly and access /ac/metrics/myapp.api.timer to view raw metrics. Prometheus scrapes the /actuator/prometheus endpoint, and you can visualize latency trends in Grafana or the Prometheus UI.