Prometheus Reporter for Telemetry.Metrics definitions.

Provide a list of metric definitions to the init/2 function. It's recommended to run TelemetryMetricsPrometheus under a supervision tree, usually under Application.

def start(_type, _args) do
  # List all child processes to be supervised
  children = [
    {TelemetryMetricsPrometheus, [metrics: metrics()]}
  ...
  ]

  opts = [strategy: :one_for_one, name: ExampleApp.Supervisor]
  Supervisor.start_link(children, opts)
end

defp metrics, do:
  [
    counter("http.request.count"),
    sum("http.request.payload_size", unit: :byte),
    last_value("vm.memory.total", unit: :byte)
  ]

By default, metrics are exposed on port 9568 at /metrics. The port number can be configured if necessary. You are not required to use the included server, though it is recommended. https is not supported yet, in which case the TelemetryMetricsPrometheus.Core library should be used instead. The TelemetryMetricsPrometheus.Core.scrape/1 function will expose the metrics in the Prometheus text format.

Please see the TelemetryMetricsPrometheus.Core docs for information on metric types and units.

Telemetry Events

• [:prometheus_metrics, :plug, :stop] - invoked at the end of every scrape. The measurement returned is :duration and the metadata is the conn map for the call.

A suggested Distribution definition might look like:

Metrics.distribution("prometheus_metrics.scrape.duration.milliseconds",
  reporter_options: [buckets: [0.05, 0.1, 0.2, 0.5, 1]],
  description: "A histogram of the request duration for prometheus metrics scrape.",
  event_name: [:prometheus_metrics, :plug, :stop],
  measurement: :duration,
  tags: [:name],
  tag_values: fn %{conn: conn} ->
    %{name: conn.private[:prometheus_metrics_name]}
  end,
  unit: {:native, :millisecond}
)