Hyper方案 服务监控告警（Prometheus + Grafana）

Hyperf 官方内置 hyperf/metric，原生支持 Prometheus，零侵入自动采集。 --- 安装composerrequire hyperf/metric php bin/hyperf.php vendor:publish--id=hyperf.metric ---1. 配置 config/autoload/metric.php：<?phpreturn['default'=>env('METRIC_DRIVER','prometheus'),'use_standalone_process'=>true, // 独立进程采集，不影响业务'metric'=>['prometheus'=>['driver'=>\Hyperf\Metric\Adapter\Prometheus\MetricFactory::class,'mode'=>\Hyperf\Metric\Adapter\Prometheus\Constants::SCRAPE_MODE,'namespace'=>env('APP_NAME','hyperf'),'scrape_host'=>'0.0.0.0','scrape_port'=>9502,'scrape_path'=>'/metrics',],],];---2. 自动采集（零代码） 框架自动采集以下指标，无需任何代码： hyperf_http_requests_total# HTTP 请求总数hyperf_http_request_duration_ms# 请求耗时分布hyperf_db_query_duration_ms# DB 查询耗时hyperf_redis_commands_total# Redis 命令数hyperf_coroutine_count# 协程数量hyperf_worker_memory_bytes# Worker 内存---3. 业务自定义指标<?php namespace App\Service;use Hyperf\Di\Annotation\Inject;use Hyperf\Metric\Contract\MetricFactoryInterface;class OrderService{#[Inject]private MetricFactoryInterface$metric;publicfunctioncreateOrder(int$userId, array$items): array{$timer=$this->metric->makeHistogram(name:'order_create_duration_ms', labelNames:['status'],)->startTimer();try{$order=Order::create(['user_id'=>$userId,'total'=>collect($items)->sum('price'),]);// 订单计数$this->metric->makeCounter('orders_total',['status'])->with('success')->add(1);// 订单金额分布$this->metric->makeHistogram('order_amount',['currency'])->with('CNY')->observe($order->total);return$order->toArray();}catch(\Throwable$e){$this->metric->makeCounter('orders_total',['status'])->with('failed')->add(1);throw$e;}finally{$timer->end(['status'=>isset($order)?'success':'failed']);}}}---4. 告警中间件（慢请求 + 错误率）<?php namespace App\Middleware;use Hyperf\Metric\Contract\MetricFactoryInterface;use Psr\Http\Message\ResponseInterface;use Psr\Http\Message\ServerRequestInterface;use Psr\Http\Server\MiddlewareInterface;use Psr\Http\Server\RequestHandlerInterface;class MetricMiddleware implements MiddlewareInterface{publicfunction__construct(private MetricFactoryInterface$metric){}publicfunctionprocess(ServerRequestInterface$request, RequestHandlerInterface$handler): ResponseInterface{$start=microtime(true);$path=$request->getUri()->getPath();try{$response=$handler->handle($request);$status=$response->getStatusCode();}catch(\Throwable$e){$status=500;throw$e;}finally{$duration=(microtime(true)-$start)*1000;$this->metric->makeHistogram('http_duration_ms',['path','status'])->with($path,(string)$status)->observe($duration);$this->metric->makeCounter('http_requests_total',['path','status'])->with($path,(string)$status)->add(1);}return$response;}}---5. Docker Compose（完整监控栈） services: prometheus: image: prom/prometheus:latest volumes: - ./prometheus.yml:/etc/prometheus/prometheus.yml ports: -"9090:9090"grafana: image: grafana/grafana:latest ports: -"3000:3000"environment: GF_SECURITY_ADMIN_PASSWORD: admin volumes: - grafana_data:/var/lib/grafana alertmanager: image: prom/alertmanager:latest volumes: - ./alertmanager.yml:/etc/alertmanager/alertmanager.yml ports: -"9093:9093"volumes: grafana_data: prometheus.yml： global: scrape_interval: 15s scrape_configs: - job_name: hyperf static_configs: - targets:['host.docker.internal:9502']rule_files: - alerts.yml alerting: alertmanagers: - static_configs: - targets:['alertmanager:9093']---6. 告警规则 alerts.yml： groups: - name: hyperf rules:# 错误率 > 5%- alert: HighErrorRate expr:|rate(http_requests_total{status=~"5.."}[5m])/ rate(http_requests_total[5m])>0.05for: 2m labels: severity: critical annotations: summary:"Error rate {{$value| humanizePercentage }}"# P99 延迟 > 1s- alert: SlowResponse expr:|histogram_quantile(0.99, rate(http_duration_ms_bucket[5m]))>1000for: 5m labels: severity: warning annotations: summary:"P99 latency {{$value}}ms"# 内存 > 512MB- alert: HighMemory expr: hyperf_worker_memory_bytes>536870912for: 5m labels: severity: warning alertmanager.yml（钉钉/企微通知）： route: receiver: dingtalk receivers: - name: dingtalk webhook_configs: - url:'http://dingtalk-webhook/dingtalk/robot1/send'send_resolved:true--- 核心要点： - use_standalone_process:true独立进程暴露 /metrics，不占用业务端口 - 框架自动采集覆盖80% 场景，只需对核心业务补充自定义指标 - 告警阈值：错误率5%、P99>1s、内存>512MB 是生产常用起点 - Grafana 直接导入 Dashboard ID11316（Hyperf 社区模板）