Awesome

This library provides centralized metrics collection across modules. It also provides a Prometheus / OpenMetrics compatible HTTP server exposing the collected metrics.

Collecting metrics

The core component of the metrics is the metrics.Collector interface. You can create a new instance of this interface by calling metrics.New() with a GeoIP lookup provider from the geoip library as a parameter. You can then dynamically create metrics:

m := metrics.New(geoip)
testCounter, err := m.CreateCounter(
    "test", // Name of the metric
    "MB", // Unit of the metric
    "This is a test", // Help text of the metric
)

You can then increment the counter:

testCounter.Increment()
testCounter.IncrementBy(5)

Alternatively, you can also create a CounterGeo to make a label automatically based on GeoIP lookup:

testCounter, err := m.CreateCounterGeo(
    "test", // Name of the metric
    "MB", // Unit of the metric
    "This is a test", // Help text of the metric
)
testCounter.Increment(net.ParseIP("127.0.0.1"))

If you need a metric that can be decremented or set directly you can use the Gauge type instead. Each Create* method also has a counterpart named MustCreate*, which panics instead of returning an error.

Custom labels

Each of the metric methods allow adding extra labels:

testCounter.Increment(
    net.ParseIP("127.0.0.1"),
    metrics.Label("foo", "bar"),
    metrics.Label("somelabel","somevalue")
)

The following rules apply and will cause a panic if violated:

• Label names and values cannot be empty.
• The country label name is reserved for GeoIP usage.

The metrics also have a WithLabels() method that allow for creating a copy of a metric already primed with a set of labels. This can be used when passing metrics to other modules that need to be scoped.

Using the metrics server

The metrics server exposes the collected metrics on an HTTP webserver in the Prometheus / OpenMetrics format. It requires the service library and a logger from the log library to work properly:

server := metrics.NewServer(
    metrics.Config{
        ServerConfiguration: http.ServerConfiguration{
            Listen:       "127.0.0.1:8080",
        },
        Enable:              true,
        Path:                "/metrics",
    },
    metricsCollector,
    logger,
)

lifecycle := service.NewLifecycle(server)
go func() {
    if err := lifecycle.Run(); err != nil {
        // Handle crash
    } 	
}()

//Later:
lifecycle.Stop(context.Background())

Alternatively, you can skip the full HTTP server and request a handler that you can embed in any Go HTTP server:

handler := metrics.NewHandler(
    "/metrics",
    metricsCollector
)
http.ListenAndServe("0.0.0.0:8080", handler)