Module to create an HTTP server to expose Prometheus metrics.
By default it:
- runs an HTTP server to expose metrics
- instantiates and returns a prom-client
- patches http server to get request response time
- allows to define custom metrics through prom-client
Supported and tested : >= 14.0
Version | Supported | Tested |
---|---|---|
16.x | yes | yes |
18.x | yes | yes |
20.x | yes | yes |
It works with different HTTP servers:
By default you can access your metrics on this endpoint: http://localhost:9350/metrics
In order to handle remote attack and to avoid high cardinality, the default behavior of this lib is to
override the path
label in case of HTTP code 404.
$ npm install prometheus-middleware --save
const APM = require('prometheus-middleware')
const apm = new APM()
apm.init()
const APM = require('prometheus-middleware')
const apm = new APM()
apm.init()
// ----------------
const counter = new apm.client.Counter({
name: 'metric_name',
help: 'metric_help',
})
// ----------------
counter.inc(); // Increment by 1
counter.inc(10); // Increment by 10
The metrics system is exactly the same as in prom-client librairy.
The config is an JSON object which accepts the following property:
const APM = require('prometheus-middleware')
const apm = new APM({
METRICS_ROUTE: "/metrics",
PORT: 9350,
PROM_CLIENT_CONF: {},
HTTP_DURATION_BUCKETS: [0.01, 0.02, 0.05, 0.1, 0.2, 0.5, 1, 2]
HTTP_SUMMARY_PERCENTILES: [0.5, 0.9, 0.95, 0.99]
NORMALIZE_ENDPOINT: false,
})
apm.init()
If you use a graceful exit method or if you want to simply stop the APM, you should run:
const APM = require('prometheus-middleware')
const apm = new APM()
apm.init()
process.on('SIGTERM', () => {
apm.destroy()
})
Property | Default | Description |
---|---|---|
METRICS_ROUTE | /metrics | Route to expose the metrics |
PORT | 9350 | Port to listen metrics |
PROM_CLIENT_CONF | {} | Configuration of the prom-client lib |
HTTP_DURATION_BUCKETS | [0.01, 0.02, 0.05, 0.1, 0.2, 0.5, 1, 2] | List of values for HTTP request duration |
HTTP_SUMMARY_PERCENTILES | [0.5, 0.9, 0.95, 0.99] | List of values for HTTP request percentiles |
NORMALIZE_ENDPOINT | true | Normalize endpoint by occulting ids, and query parameters |
To see how to use the module you can refer to the example folder.
The agent use debug module in order not to pollute your logs. If you want to see all agent output just use DEBUG environment variable:
DEBUG=prometheus-middleware* node myApp.js
$ npm test
Coverage report can be found in coverage/.