Prometheus Client Integration

The go-metrics package supports Prometheus client integration for exposing metrics in a Prometheus-compatible format.

Configuration

The Prometheus client integration is enabled by using the prometheus namespace in the go-metrics configuration. This namespace provides a Registry interface that can be used to register metrics with Prometheus.

Example:

import (
              "github.com/docker/go-metrics"
              "github.com/prometheus/client_golang/prometheus"
          )
          
          func main() {
              // Create a Prometheus registry.
              registry := prometheus.NewRegistry()
          
              // Create a new go-metrics registry with the Prometheus namespace.
              metricsRegistry := metrics.NewRegistry(metrics.PrometheusNamespace(registry))
          
              // Register a counter metric.
              metricsRegistry.GetOrRegisterCounter("requests", "total number of requests").Inc(1)
          
              // Run the Prometheus server.
              http.Handle("/metrics", prometheus.Handler())
              http.ListenAndServe(":9090", nil)
          }
          

References:

• metrics/prometheus/prometheus.go

Metrics Registration

The go-metrics package provides various methods to register different types of metrics with Prometheus, including:

• Counters: Count occurrences of an event.
• Gauges: Measure a specific value.
• Timers: Measure the duration of an event.
• Histograms: Measure the distribution of values.

To register a metric, use the GetOrRegister method on the prometheus.Registry interface, passing in the metric type, name, and help text.

Example:

// Register a counter metric.
          metricsRegistry.GetOrRegisterCounter("requests", "total number of requests").Inc(1)
          
          // Register a gauge metric.
          metricsRegistry.GetOrRegisterGauge("memory_usage", "current memory usage").Update(100)
          
          // Register a timer metric.
          metricsRegistry.GetOrRegisterTimer("request_duration", "duration of requests").Update(time.Second * 2)
          
          // Register a histogram metric.
          metricsRegistry.GetOrRegisterHistogram("request_size", "distribution of request sizes").Update(1024)
          

References:

• metrics/prometheus/prometheus.go

Metric Naming Conventions

The go-metrics package follows the Prometheus metric naming conventions:

• Metric names should be lowercase and separated by underscores.
• Metric names should be descriptive.
• Metric labels should be lowercase and separated by underscores.
• Metric labels should be unique within a metric.

References:

• https://prometheus.io/docs/concepts/metric_types/

Metric Collection

The Prometheus client integration exposes metrics as Prometheus metrics. The prometheus.Registry interface provides a handler that can be used to serve metrics to Prometheus.

Example:

// Run the Prometheus server.
          http.Handle("/metrics", prometheus.Handler())
          http.ListenAndServe(":9090", nil)
          

References:

• https://github.com/prometheus/client_golang/blob/master/prometheus/handler.go

Metric Access

You can directly access the registered metrics through the prometheus.Registry interface. For example, you can retrieve a specific metric and update its value.

Example:

// Get the registered counter metric.
          counter := metricsRegistry.GetOrRegisterCounter("requests", "total number of requests")
          
          // Increment the counter value.
          counter.Inc(1)
          

References:

• metrics/prometheus/prometheus.go
• https://github.com/prometheus/client_golang/blob/master/prometheus/prometheus.go