指标监控插件

go-wind-plugins 提供统一的指标采集和导出接口，支持 Prometheus、Datadog 等。

一、Metrics 接口

type Metrics interface {
    Counter(name string, tags ...Tag) Counter
    Gauge(name string, tags ...Tag) Gauge
    Histogram(name string, tags ...Tag) Histogram
    Timer(name string, tags ...Tag) Timer
}

二、适配器列表

适配器导入路径后端
Prometheusplugins/metrics/prometheusPrometheus + Grafana
Datadogplugins/metrics/datadogDatadog Agent
StatsDplugins/metrics/statsdStatsD / DogStatsD
OTLPplugins/metrics/otlpOpenTelemetry Collector

三、Prometheus

import _ "github.com/tx7do/go-wind-plugins/metrics/prometheus"

YAML 配置

metrics:
  prometheus:
    enabled: true
    path: /metrics          # 采集路径
    namespace: myapp        # 指标前缀
    subsystem: api
    labels:
      service: my-service
      env: production

自定义指标

import "github.com/tx7do/go-wind/metrics"

// Counter（计数器）
requestCount := metrics.Counter("requests_total",
    metrics.Tag("method", "GET"),
    metrics.Tag("path", "/api/users"),
)
requestCount.Inc()

// Gauge（瞬时值）
activeConnections := metrics.Gauge("active_connections")
activeConnections.Set(42)
activeConnections.Inc()
activeConnections.Dec()

// Histogram（直方图）
latency := metrics.Histogram("request_duration_seconds",
    metrics.Tag("path", "/api/users"),
)
latency.Observe(0.123)  // 记录耗时

// Timer（计时器）
timer := metrics.Timer("db_query_duration")
timer.Start()
// ... 执行数据库查询
timer.Stop()

暴露 /metrics 端点

Prometheus 插件自动在 HTTP Server 上注册 /metrics 路径：

// 无需额外代码
// Prometheus 插件 init() 自动注册 /metrics handler
// Prometheus Server 通过 http://localhost:8080/metrics 采集

Prometheus 采集配置

# prometheus.yml
scrape_configs:
  - job_name: 'my-service'
    scrape_interval: 15s
    static_configs:
      - targets: ['localhost:8080']

四、Datadog

import _ "github.com/tx7do/go-wind-plugins/metrics/datadog"

YAML 配置

metrics:
  datadog:
    enabled: true
    addr: "localhost:8125"       # DogStatsD 地址
    namespace: myapp
    tags:
      - "service:my-service"
      - "env:production"
    flush_interval: 10s

五、常用指标模式

5.1 HTTP 请求监控

func MetricsMiddleware(next http.Handler) http.Handler {
    return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
        start := time.Now()

        next.ServeHTTP(w, r)

        duration := time.Since(start).Seconds()
        status := strconv.Itoa(w.(*responseWriter).statusCode)

        metrics.Counter("http_requests_total",
            metrics.Tag("method", r.Method),
            metrics.Tag("path", r.URL.Path),
            metrics.Tag("status", status),
        ).Inc()

        metrics.Histogram("http_request_duration_seconds",
            metrics.Tag("method", r.Method),
            metrics.Tag("path", r.URL.Path),
        ).Observe(duration)
    })
}

5.2 业务指标

// 订单指标
metrics.Counter("orders_created_total",
    metrics.Tag("type", "vip"),
).Inc()

// 支付金额
metrics.Histogram("payment_amount",
    metrics.Tag("currency", "CNY"),
).Observe(99.9)

// 队列积压
metrics.Gauge("queue_size",
    metrics.Tag("queue", "order_events"),
).Set(float64(queue.Len()))

六、Grafana 面板

推荐 Grafana 面板配置：

面板PromQL类型
QPSrate(http_requests_total[5m])Graph
延迟 P99histogram_quantile(0.99, rate(http_request_duration_seconds_bucket[5m]))Graph
错误率rate(http_requests_total{status=~"5.."}[5m]) / rate(http_requests_total[5m])Graph
活跃连接active_connectionsGauge

相关文档