Skip to content

README.md

Latest commit

 

History

History
91 lines (64 loc) · 2.51 KB

README.md

File metadata and controls

91 lines (64 loc) · 2.51 KB
Raw

avaje-metrics-prometheus

Exports avaje-metrics in Prometheus text exposition format.

This module is a lightweight pull/scrape exporter. It does not run a scheduler and does not start an HTTP server; expose the scrape body from the web framework already used by the application.

Maven dependency

<dependency>
  <groupId>io.avaje</groupId>
  <artifactId>avaje-metrics-prometheus</artifactId>
  <version>${version}</version>
</dependency>

If the application uses module-info.java, also add:

requires io.avaje.metrics.prometheus;

Basic usage

import io.avaje.metrics.Metrics;
import io.avaje.metrics.prometheus.PrometheusMetrics;

var prometheus = PrometheusMetrics.builder()
  .registry(Metrics.registry())
  .build();

String body = prometheus.scrape();

For an HTTP endpoint, return:

  • body: prometheus.scrape()
  • content type: PrometheusMetrics.CONTENT_TYPE

You can also stream to an Appendable:

prometheus.write(writer);

Builder options

var prometheus = PrometheusMetrics.builder()

  // Use a non-default MetricRegistry (defaults to Metrics.registry())
  .registry(myRegistry)

  // Advanced: skip timer metrics whose cumulative total duration is below this threshold.
  // This is usually less useful for Prometheus because scrapes use cumulative values.
  .timedThresholdMicros(1_000)

  // Include scrape-window max gauges for timers and meters (disabled by default).
  .includeMax(true)

  .build();

Metric mapping

avaje metrics are collected using CollectionMode.CUMULATIVE and mapped as follows:

avaje type Prometheus output
Counter name_total counter
Timer name_seconds summary with _count and _sum samples
bucketed Timer name_seconds histogram with _bucket, _count, and _sum samples
Meter name_count_total and name_total counters
GaugeLong / GaugeDouble name gauge

max values are disabled by default because avaje cumulative collection still treats max as a scrape-window value. Enable .includeMax(true) only when that behavior is useful.

timedThresholdMicros(...) is available as an advanced option, but it applies to the cumulative process-lifetime timer total, so it is usually less useful for Prometheus scraping than it is for delta/scheduled reporters.

More docs