moov-io/base/admin

Package admin implements an http.Server which can be used for operations and monitoring tools. It's designed to be shipped (and ran) inside an existing Go service.

Here's an example of adding admin.Server to serve Prometheus metrics:

import (
    "fmt"
    "os"

    "github.com/moov-io/base/admin"

    "github.com/go-kit/log"
)

var logger log.Logger

// in main.go or cmd/server/main.go

adminServer, err := admin.New(Opts{
	Addr: ":9090",
})
if err != nil {
    // handle error
}
go func() {
	logger.Log("admin", fmt.Sprintf("listening on %s", adminServer.BindAddr()))
	if err := adminServer.Listen(); err != nil {
		err = fmt.Errorf("problem starting admin http: %v", err)
		logger.Log("admin", err)
		// errs <- err // send err to shutdown channel
	}
}()
defer adminServer.Shutdown()

Endpoints

An Admin server has some default endpoints that are useful for operational support and monitoring.

Liveness Probe

This endpoint inspects a set of liveness functions and returns 200 OK if all functions return without errors. If errors are found then a 400 Bad Request response with a JSON object is returned describing the errors.

GET /live

Liveness probes can be registered with the following callback:

func (s *Server) AddLivenessCheck(name string, f func() error)

Readiness Probe

This endpoint inspects a set of readiness functions and returns 200 OK if all functions return without errors. If errors are found then a 400 Bad Request response with a JSON object is returned describing the errors.

GET /ready

Readiness probes can be registered with the following callback:

func (s *Server) AddReadinessCheck(name string, f func() error)

Metrics

This endpoint returns prometheus metrics registered to the prometheus/client_golang singleton metrics registry. Their promauto package can be used to add Counters, Guages, Histograms, etc. The default Go metrics provided by prometheus/client_golang are included.

GET /metrics

...
# HELP promhttp_metric_handler_requests_total Total number of scrapes by HTTP status code.
# TYPE promhttp_metric_handler_requests_total counter
promhttp_metric_handler_requests_total{code="200"} 0
promhttp_metric_handler_requests_total{code="500"} 0
promhttp_metric_handler_requests_total{code="503"} 0
# HELP stream_file_processing_errors Counter of stream submitted ACH files that failed processing
# TYPE stream_file_processing_errors counter
stream_file_processing_errors 0