使用 prometheus 监控应用

const Koa = require('koa')
const app = new Koa()

app.use(async (ctx) => {
  const r = Math.random() * 100
  await sleep(r)
  if (r > 90) {
    ctx.status = 400
  }
  ctx.body = r
})

app.listen(3000)

const { register, Counter } = require('prom-client')

// 建立一个 counter 统计指标
const counter = new Counter({
  name: 'http_requests_total',
  help: 'Counter for total requests received',
  labelNames: ['path', 'method', 'status'], // labels 为了统计分组, 可以区分不同维度的指标
})

// 监控指标中间件
app.use(async (ctx, next) => {
  // 路径 /metrics 为监控指标 route, 返回指标
  if (ctx.path === '/metrics') {
    ctx.set('Content-Type', register.contentType)
    ctx.body = register.metrics()
    return
  }

  // 其他路由均统计指标
  try {
    await next()
  } finally {
    const labels = {
      path: ctx.path,
      method: ctx.method,
      status: ctx.status,
    }
    counter.inc(labels, 1)
  }
})

const hrtime2ms = (hrtime) => (hrtime[0] * 1e9 + hrtime[1]) / 1e6 // hrtime 转化成 ms

const histogram = new Histogram({
  name: `http_request_duration_ms`,
  help: 'Duration of HTTP requests in ms',
  labelNames: ['path', 'method', 'status'],
  buckets: [5, 10, 25, 50, 100, 250, 500, 1000], // buckets 单位 ms
})

// 监控指标中间件
app.use(async (ctx, next) => {
  // 路径 /metrics 为监控指标 route, 返回指标
  if (ctx.path === '/metrics') {
    ctx.set('Content-Type', register.contentType)
    ctx.body = register.metrics()
    return
  }

  // 其他路由均统计指标
  const start = process.hrtime() // 开始时间
  try {
    await next()
  } finally {
    const dur = hrtime2ms(process.hrtime(start)) // 计算请求处理时间
    const labels = {
      path: ctx.path,
      method: ctx.method,
      status: ctx.status,
    }
    histogram.observe(labels, dur) // 统计响应时间
  }
})

git clone https://github.com/zcong1993/prometheus-grafana.git

# my global config# my global config
global:
  scrape_interval: 5s # 每 5 秒抓取一次指标, 仅做测试, 生产环境不应这么频繁
  evaluation_interval: 15s # Evaluate rules every 15 seconds. The default is every 1 minute.
  # scrape_timeout is set to the global default (10s).

  # Attach these labels to any time series or alerts when communicating with
  # external systems (federation, remote storage, Alertmanager).
  external_labels:
    monitor: 'codelab-monitor'

# Load rules once and periodically evaluate them according to the global 'evaluation_interval'.
rule_files:
  # - "first.rules"
  # - "second.rules"

# A scrape configuration containing exactly one endpoint to scrape:
# Here it's Prometheus itself.
scrape_configs:
  # The job name is added as a label `job=<job_name>` to any timeseries scraped from this config.
  - job_name: 'prometheus'

    # metrics_path defaults to '/metrics'
    # scheme defaults to 'http'.

    static_configs:
      - targets: ['localhost:9090']
  # Add your own jobs here.
  - job_name: 'koa-app'
    static_configs:
      - targets: ['你的本机 IP:3000'] # 由于 prometheus 是从 docker 内访问本机, 所以不能使用 localhost

sum(rate(http_requests_total{job=~"koa-app", path=~".*"}[2m])) by (path)

avg(increase(http_request_duration_ms_sum{job=~"koa-app", path=~".*"}[1m]) / increase(http_request_duration_ms_count{job=~"koa-app", path=~".*"}[1m]) >0) by (path)

histogram_quantile(0.90, sum(irate(http_request_duration_ms_bucket{job=~"koa-app", path=~".*"}[1m])) by (path, le))

sum(irate(http_requests_total{status!~"200",job=~"koa-app", path=~".*"}[5m])) BY (job, path, status) / IGNORING(status) GROUP_LEFT() sum(irate(http_requests_total{job=~"koa-app", path=~".*"}[5m])) BY (job, path) * 100