express-prometheus-middleware v1.2.0

Express Prometheus Middleware

This is a middleware for express servers, that expose metrics for prometheus.

The metrics exposed allows to calculate common RED (Request, Error rate, Duration of requests), and USE (Utilisation, Error rate, and Saturation), metrics

Install

yarn add express-prometheus-middleware
# or
npm i --save express-prometheus-middleware

Usage

Options

Example

const express = require('express');
const promMid = require('express-prometheus-middleware');
const app = express();

const PORT = 9091;
app.use(promMid({
  metricsPath: '/metrics',
  collectDefaultMetrics: true,
  requestDurationBuckets: [0.1, 0.5, 1, 1.5],
  requestLengthBuckets: [512, 1024, 5120, 10240, 51200, 102400],
  responseLengthBuckets: [512, 1024, 5120, 10240, 51200, 102400],
  /**
   * Uncomenting the `authenticate` callback will make the `metricsPath` route
   * require authentication. This authentication callback can make a simple
   * basic auth test, or even query a remote server to validate access.
   * To access /metrics you could do:
   * curl -X GET user:password@localhost:9091/metrics
   */
  // authenticate: req => req.headers.authorization === 'Basic dXNlcjpwYXNzd29yZA==',
  /**
   * Uncommenting the `extraMasks` config will use the list of regexes to
   * reformat URL path names and replace the values found with a placeholder value
  */
  // extraMasks: [/..:..:..:..:..:../],
  /**
   * The prefix option will cause all metrics to have the given prefix.
   * E.g.: `app_prefix_http_requests_total`
   */
  // prefix: 'app_prefix_',
  /**
   * Can add custom labels with customLabels and transformLabels options
   */
  // customLabels: ['contentType'],
  // transformLabels(labels, req) {
  //   // eslint-disable-next-line no-param-reassign
  //   labels.contentType = req.headers['content-type'];
  // },
}));

// curl -X GET localhost:9091/hello?name=Chuck%20Norris
app.get('/hello', (req, res) => {
  console.log('GET /hello');
  const { name = 'Anon' } = req.query;
  res.json({ message: `Hello, ${name}!` });
});

app.listen(PORT, () => {
  console.log(`Example api is listening on http://localhost:${PORT}`);
});

Metrics exposed

• Default metrics from prom-client
• http_requests_total: Counter for total requests received, has labels route, method, status
• http_request_duration_seconds: - Duration of HTTP requests in seconds, has labels route, method, status

The labels route and status are normalized:

• route: will normalize id like route params
• status: will normalize to status code family groups, like 2XX or 4XX.

Example prometheus queries

In the examples below, Suppose you tagged your application as "myapp", in the prometheus scrapping config.

Running instances

sum(up{app="myapp"})

Overall error rate

Rate of http status code 5XX responses

sum(rate(http_requests_total{status="5XX", app="myapp"}[5m]))

95% of requests served within seconds

histogram_quantile(0.95, sum(rate(http_request_duration_seconds_bucket{app="myapp"}[5m])) by (le))

95% of request length

histogram_quantile(0.95, sum(rate(http_request_length_bytes_bucket{app="myapp"}[5m])) by (le))

95% of response length

histogram_quantile(0.95, sum(rate(http_response_length_bytes_bucket{app="myapp"}[5m])) by (le))

Average response time in seconds

sum(rate(http_request_duration_seconds_sum{app="myapp"}[5m])) by (instance) / sum(rate(http_request_duration_seconds_count{app="myapp"}[5m])) by (instance)

Overall Request rate

sum(rate(http_requests_total{app="myapp"}[5m])) by (instance)

Request rate by route

In this example we are removing some health/status-check routes, replace them with your needs.

sum(rate(http_requests_total{app="myapp", route!~"/|/healthz"}[5m])) by (instance, route)

CPU usage

rate(process_cpu_system_seconds_total{app="myapp"}[5m])
rate(process_cpu_user_seconds_total{app="myapp"}[5m])

Memory usage

nodejs_heap_size_total_bytes{app="myapp"}
nodejs_heap_size_used_bytes{app="myapp"}