prometheus_client

Pub Version Dart CI

This is a simple Dart implementation of the Prometheus client library, similar to to libraries for other languages. It supports the default metric types like gauges, counters, summaries, or histograms. Metrics can be exported using the text format. To expose them in your server application the package comes with the package prometheus_client_shelf to integrate metrics with a shelf handler. In addition, it comes with some plug-in ready metrics for the Dart runtime.

You can find the latest updates in the changelog.

Usage

A simple usage example:

import 'dart:io';

import 'package:prometheus_client/format.dart' as format;
import 'package:prometheus_client/prometheus_client.dart';
import 'package:prometheus_client/runtime_metrics.dart' as runtime_metrics;

main() async {
  // Register runtime metrics with the default metrics registry
  runtime_metrics.register();

  // Create a Histogram metrics without labels. Always register your metric,
  // either at the default registry or a custom one.
  final durationHistogram = Histogram(
    name: 'http_request_duration_seconds',
    help: 'The duration of http requests in seconds.',
  )
    ..register();

  // Create a metric of type counter, with a label for the requested path:
  final metricRequestsCounter = Counter(
    name: 'metric_requests_total',
    help: 'The total amount of requests of the metrics.',
    labelNames: ['path'],
  )
    ..register();

  // Create a http server
  final server = await HttpServer.bind(
    InternetAddress.loopbackIPv4,
    8080,
  );
  print('Listening on localhost:${server.port}');

  await for (HttpRequest request in server) {
    // Measure the request duration
    await durationHistogram.observeDuration(Future(() async {
      // Count calls to the metric endpoint by path.
      metricRequestsCounter.labels([request.uri.path]).inc();

      // Output metrics in the text representation
      request.response.headers.add('content-type', format.contentType);
      final metrics = await CollectorRegistry.defaultRegistry.collectMetricFamilySamples();
      format.write004(request.response, metrics);
      await request.response.close();
    }));
  }
}

Start the example application and access the exposed metrics at http://localhost:8080/. For a full usage example, take a look at example/prometheus_client_example.dart.

Metrics

The package provides different kinds of metrics:

Counter

Counter is a monotonically increasing counter. Counters can only be incremented, either by one or a custom amount:

final requestsCounter = Counter(
  name: 'metric_requests_total',
  help: 'The total amount of requests of the metrics.',
);

requestsCounter.inc();
requestsCounter.inc(64.0);

Gauge

A Gauge represents a value that can go up and down:

final gauge = Gauge(
  name: 'my_metric',
  help: 'Help!',
);

gauge.value = 1337.0;
gauge.inc(2.0);
gauge.dec(4.0);

Histogram

Histogram allows aggregatable distributions of events, such as request latencies. The buckets of the histogram are customizable and support both linear and exponential buckets.

final histogram = Histogram(
    name: 'my_metric',
    help: 'Help!',
);

histogram.observe(20.0);
await histogram.observeDuration(() async {
  // Some code
});
histogram.observeDurationSync(() {
  // Some code
});

Summary

Similar to a Histogram, a Summary samples observations (usually things like request durations and response sizes). While it also provides a total count of observations and a sum of all observed values, it calculates configurable quantiles over a sliding time window.

final summary = Summary(
  name: 'my_metric',
  help: 'Help!',
);

summary.observe(20.0);
await summary.observeDuration(() async {
  // Some code
});
summary.observeDurationSync(() {
  // Some code
});

Labels

Metrics can optionally have labels. You can pass label names during metrics creation. Later on, you can access a child metric for your label values via the labels() function.

final requestsCounter = Counter(
  name: 'metric_requests_total',
  help: 'The total amount of requests of the metrics.',
  labelNames: ['path'],
);

requestsCounter.labels(['my/path/']).inc();

Collect

For metrics that should provide the current value during scraping, each metric type provides a collectCallback that is executed during scraping. You can use the callback to update the metric while scraping. The callback can either be sync, or return a Future:

final gauge = Gauge(
  name: 'my_metric',
  help: 'Help!',
  collectCallback: (gauge) {
    // Callback is executed every time the metric is collected.
    // Use the function parameter to access the gauge and update
    // its value.
    gauge.value = 1337;
  },
);

Default Metrics

The RuntimeCollector provides a basic set of built-in metrics. However, only a limited set of the recommended standard metrics is implemented, as Dart doesn't expose them.

Outputting Metrics

The package comes with function for serializing the metrics into the Prometheus metrics text format:

final buffer = StringBuffer();
final metrics = await CollectorRegistry.defaultRegistry.collectMetricFamilySamples();
format.write004(buffer, metrics);
print(buffer.toString());

If you are using shelf, take a look at prometheus_client_shelf for providing a metrics endpoint.

Features and bugs

Please file feature requests and bugs at the issue tracker.

Libraries

format
A library to export metrics in the Prometheus text representation.
prometheus_client
A library containing the core elements of the Prometheus client, like the CollectorRegistry and different types of metrics like Counter, Gauge and Histogram.
runtime_metrics
A library exposing metrics of the Dart runtime.