Provides a bridge between Finagle and Prometheus metrics
Clone or download


Build Status Coverage Status

finagle-prometheus is a library that provides a bridge between Finagle and Prometheus. It contains everything you need to get metrics from a Finagle application into Prometheus.

Getting started

The library is available in Maven Central, for both Scala 2.11 and 2.12, so just add the following dependency:

"com.samstarling" %% "finagle-prometheus" % "0.0.8"


Exposing metrics

The MetricsService exposes all metrics over HTTP in the format that Prometheus expects when collecting metrics. All it needs is a CollectorRegistry:

val registry = CollectorRegistry.defaultRegistry
val metricsService = new MetricsService(registry)

You can then mount this service alongside your other HTTP handlers, and configure Prometheus to scrape that endpoint.

Existing Finagle metrics

Finagle exposes a lot of metrics internally. They are reported to a default StatsReceiver. You can replace this with the PrometheusStatsReceiver, which will transform the Finagle metrics into Prometheus metrics, and register them with the given CollectorRegistry.

Create one like this:

val registry = CollectorRegistry.defaultRegistry
val statsReceiver = new PrometheusStatsReceiver(registry)

And use it (for example) when building a Client:

    .newService("", "GitHub")

The best thing you can do is use the PrometheusStatsReceiver by default. To do this, create a file at resources/META-INF/services/com.twitter.finagle.stats.StatsReceiver with the following contents:


The default constructor for the class will be used. This behaviour is documented in Finagle's Resolver class, as well as in the util-stats documentation. For full documentation of the metrics exposed, and what they mean, see the Metrics page on the Finagle documentation site.

Custom metrics

To expose your own metrics (with labels), use the Telemetry class. You can see this in action in the CustomTelemetryService example.

val registry = CollectorRegistry.defaultRegistry
val telemetry = new Telemetry(registry, "namespace")

val counter = telemetry.counter(
    "Help text",



The library provides some existing filters that you can add to your service. At the moment, they only relate to HTTP. The metrics they export are entirely separate from existing Finagle metrics. The following filters exist:

  • HttpMonitoringFilter: gives metrics relating to the number of HTTP responses, including status code and HTTP method.
  • HttpLatencyMonitoringFilter: gives metrics relating to the latency of HTTP responses, including status code and HTTP method.

Note: your service must be of the type Service[http.Request, http.Response].