利用Micrometer集成Prometheus监控外卖API调用量与SLA达成情况-优快云博客

利用Micrometer集成Prometheus监控外卖API调用量与SLA达成情况

在高并发的外卖平台中，API的稳定性与响应性能直接决定用户体验和业务转化率。为实时掌握各接口的调用量、延迟分布及SLA（如95%请求<500ms）达成情况，需构建可观测性体系。本文基于baodanbao.com.cn.*包结构，展示如何通过Micrometer + Prometheus + Grafana实现精细化API指标采集与SLA监控。

1. 引入依赖与基础配置

在pom.xml中添加：

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

application.yml启用端点：

management:
  endpoints:
    web:
      exposure:
        include: health,info,metrics,prometheus
  metrics:
    tags:
      application: baodanbao-api
    web:
      server:
        request:
          autotime:
            enabled: true

在这里插入图片描述

2. 自定义API指标埋点

Spring Boot Actuator默认提供http.server.requests指标，但缺乏业务维度（如“是否美团订单查询”）。需手动增强。

创建拦截器记录业务标签：

package baodanbao.com.cn.interceptor;

import io.micrometer.core.instrument.MeterRegistry;
import io.micrometer.core.instrument.Timer;
import org.springframework.stereotype.Component;
import org.springframework.web.servlet.HandlerInterceptor;

import javax.servlet.http.HttpServletRequest;
import javax.servlet.http.HttpServletResponse;
import java.util.concurrent.ConcurrentHashMap;

@Component
public class ApiMetricsInterceptor implements HandlerInterceptor {

    private final MeterRegistry meterRegistry;
    private final ConcurrentHashMap<String, Timer.Sample> activeSamples = new ConcurrentHashMap<>();

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

    @Override
    public boolean preHandle(HttpServletRequest request, Object handler) {
        String uri = request.getRequestURI();
        String method = request.getMethod();
        String apiName = extractApiName(uri);
        Timer.Sample sample = Timer.start(meterRegistry);
        activeSamples.put(getTraceId(request), sample);
        return true;
    }

    @Override
    public void afterCompletion(HttpServletRequest request, HttpServletResponse response, Object handler, Exception ex) {
        String traceId = getTraceId(request);
        Timer.Sample sample = activeSamples.remove(traceId);
        if (sample == null) return;

        String uri = request.getRequestURI();
        String method = request.getMethod();
        String apiName = extractApiName(uri);
        int status = response.getStatus();

        // 添加业务标签：如是否涉及美团
        String bizSource = uri.contains("/meituan/") ? "meituan" : "internal";

        sample.stop(Timer.builder("baodanbao.api.duration")
                .tag("api", apiName)
                .tag("method", method)
                .tag("status", String.valueOf(status))
                .tag("biz_source", bizSource)
                .register(meterRegistry));
    }

    private String extractApiName(String uri) {
        if (uri.startsWith("/api/order/")) return "order_query";
        if (uri.startsWith("/api/trial/")) return "trial_apply";
        return "unknown";
    }

    private String getTraceId(HttpServletRequest request) {
        return request.getRemoteAddr() + "_" + System.nanoTime();
    }
}

注册拦截器：

package baodanbao.com.cn.config;

import baodanbao.com.cn.interceptor.ApiMetricsInterceptor;
import org.springframework.context.annotation.Configuration;
import org.springframework.web.servlet.config.annotation.InterceptorRegistry;
import org.springframework.web.servlet.config.annotation.WebMvcConfigurer;

@Configuration
public class WebConfig implements WebMvcConfigurer {

    private final ApiMetricsInterceptor apiMetricsInterceptor;

    public WebConfig(ApiMetricsInterceptor apiMetricsInterceptor) {
        this.apiMetricsInterceptor = apiMetricsInterceptor;
    }

    @Override
    public void addInterceptors(InterceptorRegistry registry) {
        registry.addInterceptor(apiMetricsInterceptor).addPathPatterns("/api/**");
    }
}

3. SLA达标率计算（以95分位为例）

Prometheus本身不存储分位数，需在Micrometer中启用直方图：

// 在Timer构建时添加SLO桶
sample.stop(Timer.builder("baodanbao.api.duration")
        .tag("api", apiName)
        .tag("biz_source", bizSource)
        .serviceLevelObjectives(
            java.time.Duration.ofMillis(200),
            java.time.Duration.ofMillis(500),
            java.time.Duration.ofMillis(1000)
        )
        .publishPercentiles(0.5, 0.95, 0.99) // 发布分位数（仅内存，不推荐生产）
        .register(meterRegistry));

更推荐使用SLO Bucket + Prometheus Histogram方式。上述.serviceLevelObjectives()会生成le（less than or equal）标签的计数器，可用于计算SLA达成率。

例如，查询“美团试吃接口500ms内完成比例”：

sum(rate(baodanbao_api_duration_seconds_bucket{api="trial_apply", biz_source="meituan", le="0.5"}[5m]))
/
sum(rate(baodanbao_api_duration_seconds_count{api="trial_apply", biz_source="meituan"}[5m]))

4. 调用量与错误率监控

总调用量：rate(baodanbao_api_duration_seconds_count[1m])
错误率（5xx）：

sum(rate(baodanbao_api_duration_seconds_count{status=~"5.."}[1m]))
/
sum(rate(baodanbao_api_duration_seconds_count[1m]))

可在Grafana中配置告警：当错误率>1%或95分位>800ms时触发。

5. 高级：按商户维度细分（可选）

若需监控重点商户（如KA客户）的API体验，可在请求头中传递X-Merchant-Id，并作为Tag注入：

String merchantId = request.getHeader("X-Merchant-Id");
if (merchantId != null && isKeyAccount(merchantId)) {
    sample.stop(Timer.builder("baodanbao.api.duration")
            .tag("merchant_id", merchantId)
            // ... other tags
            .register(meterRegistry));
}

6. 验证指标输出

访问 GET /actuator/prometheus，可见类似指标：

baodanbao_api_duration_seconds_count{api="trial_apply",biz_source="meituan",method="POST",status="200"} 1245.0
baodanbao_api_duration_seconds_sum{api="trial_apply",biz_source="meituan",method="POST",status="200"} 423.8
baodanbao_api_duration_seconds_bucket{api="trial_apply",biz_source="meituan",method="POST",status="200",le="0.5"} 1180.0
baodanbao_api_duration_seconds_bucket{api="trial_apply",biz_source="meituan",method="POST",status="200",le="+Inf"} 1245.0

本文著作权归吃喝不愁app开发者团队，转载请注明出处！