This library contains helpers for easily instrumenting your code.
Currently it only supports Prometheus
It exposes a simple wrapper around the the promhttp.Handler as well as a selection of core metrics that follow the RED Method
To instrument an http handler with the core metrics you can do the following:
package vertical
import (
"net/http"
"github.com/davyj0nes/golangmetrics"
"github.com/prometheus/client_golang/prometheus"
"github.com/someCompany/some"
)
type Handler struct {
Metrics golangmetrics.CoreMetrics
Worker some.Worker
}
func NewHandler() *Handler {
return &Handler{
Metrics: golangmetrics.NewDefaultMetrics("vertical")
Worker: some.InitWorker()
}
}
func (h *Handler) Index(w http.ResponseWriter, r *http.Request) http.HandleFunc {
return func(w http.ResponseWriter, r *http.Request) {
timer := prometheus.NewTimer(h.Metrics.RequestDuration.WithLabelValues("index"))
h.Metrics.RequestRate.WithLabelValues("index").Inc()
defer timer.ObserveDuration()
someData, err := h.Worker.GoGetSomething()
if err != nil {
h.Metrics.ErrorRate.WithLabelValues("index").Inc()
return errorHandler(w, r)
}
w.Write(someData)
}
}The library supports the 4 main metric types that enable you to expose data about your applications.
A counter, very simply, is a single numeric value that can only increase. The wording that the prometheus docs uses is as follows:
A counter is a cumulative metric that represents a single monotonically increasing counter whose value can only increase or be reset to zero on restart. For example, you can use a counter to represent the number of requests served, tasks completed, or errors.
A gauge is a single numeric value that can both go up and go down.
A histogram divides data into configurable buckets as well as an overall sum.
A summary is very similar to a histogram in that it samples data and calculates configurable quantiles over time.