syncbot/internal/manager/http.go

package manager

import (
	"context"
	"encoding/json"
	"fmt"
	"log/slog"
	"net/http"
	"strconv"
	"strings"
	"time"
)

// Handler exposes the manager over HTTP:
//
//	/healthz  the process is alive (use this for a container health check)
//	/readyz   every repo has synced successfully at least once
//	/status   JSON snapshot of every repo
//	/metrics  Prometheus text format
func (m *Manager) Handler(version string) http.Handler {
	mux := http.NewServeMux()

	mux.HandleFunc("GET /healthz", func(w http.ResponseWriter, r *http.Request) {
		writeText(w, http.StatusOK, "ok\n")
	})

	mux.HandleFunc("GET /readyz", func(w http.ResponseWriter, r *http.Request) {
		var pending []string
		for _, s := range m.Statuses() {
			if s.LastSuccess.IsZero() {
				pending = append(pending, s.Name)
			}
		}
		if len(pending) > 0 {
			writeText(w, http.StatusServiceUnavailable,
				"awaiting first successful sync: "+strings.Join(pending, ", ")+"\n")
			return
		}
		writeText(w, http.StatusOK, "ready\n")
	})

	mux.HandleFunc("GET /status", func(w http.ResponseWriter, r *http.Request) {
		statuses := m.Statuses()
		body := struct {
			Version  string   `json:"version"`
			Uptime   string   `json:"uptime"`
			Repos    []Status `json:"repos"`
			Healthy  bool     `json:"healthy"`
			Failing  int      `json:"failing"`
			Reported string   `json:"reported_at"`
		}{
			Version:  version,
			Uptime:   m.Uptime().Round(time.Second).String(),
			Repos:    statuses,
			Healthy:  true,
			Reported: time.Now().UTC().Format(time.RFC3339),
		}
		for _, s := range statuses {
			if s.Failures > 0 {
				body.Failing++
				body.Healthy = false
			}
		}
		w.Header().Set("Content-Type", "application/json; charset=utf-8")
		enc := json.NewEncoder(w)
		enc.SetIndent("", "  ")
		_ = enc.Encode(body)
	})

	mux.HandleFunc("GET /metrics", func(w http.ResponseWriter, r *http.Request) {
		w.Header().Set("Content-Type", "text/plain; version=0.0.4; charset=utf-8")
		var b strings.Builder
		fmt.Fprintf(&b, "# HELP syncbot_build_info Version of the running binary.\n")
		fmt.Fprintf(&b, "# TYPE syncbot_build_info gauge\n")
		fmt.Fprintf(&b, "syncbot_build_info{version=%q} 1\n", version)
		fmt.Fprintf(&b, "# HELP syncbot_uptime_seconds Time since start.\n")
		fmt.Fprintf(&b, "# TYPE syncbot_uptime_seconds gauge\n")
		fmt.Fprintf(&b, "syncbot_uptime_seconds %s\n", seconds(m.Uptime()))

		metric(&b, "syncbot_sync_total", "counter", "Sync cycles started.")
		for _, s := range m.Statuses() {
			fmt.Fprintf(&b, "syncbot_sync_total{repo=%q} %d\n", s.Name, s.Syncs)
		}
		metric(&b, "syncbot_push_total", "counter", "Cycles that pushed to dst.")
		for _, s := range m.Statuses() {
			fmt.Fprintf(&b, "syncbot_push_total{repo=%q} %d\n", s.Name, s.Pushes)
		}
		metric(&b, "syncbot_consecutive_failures", "gauge", "Failed cycles since the last success.")
		for _, s := range m.Statuses() {
			fmt.Fprintf(&b, "syncbot_consecutive_failures{repo=%q} %d\n", s.Name, s.Failures)
		}
		metric(&b, "syncbot_refs", "gauge", "Mirrored refs.")
		for _, s := range m.Statuses() {
			fmt.Fprintf(&b, "syncbot_refs{repo=%q} %d\n", s.Name, s.Refs)
		}
		metric(&b, "syncbot_last_success_timestamp_seconds", "gauge", "Unix time of the last successful sync.")
		for _, s := range m.Statuses() {
			fmt.Fprintf(&b, "syncbot_last_success_timestamp_seconds{repo=%q} %d\n", s.Name, unix(s.LastSuccess))
		}
		metric(&b, "syncbot_last_duration_seconds", "gauge", "Duration of the last sync cycle.")
		for _, s := range m.Statuses() {
			fmt.Fprintf(&b, "syncbot_last_duration_seconds{repo=%q} %s\n", s.Name,
				strconv.FormatFloat(float64(s.LastDurMS)/1000, 'f', 3, 64))
		}
		_, _ = w.Write([]byte(b.String()))
	})

	return mux
}

// Serve runs the HTTP endpoint until ctx is cancelled.
func (m *Manager) Serve(ctx context.Context, addr, version string, log *slog.Logger) error {
	srv := &http.Server{
		Addr:              addr,
		Handler:           m.Handler(version),
		ReadHeaderTimeout: 10 * time.Second,
	}
	go func() {
		<-ctx.Done()
		shutdown, cancel := context.WithTimeout(context.Background(), 5*time.Second)
		defer cancel()
		_ = srv.Shutdown(shutdown)
	}()

	log.Info("http listening", "addr", addr,
		"endpoints", "/healthz /readyz /status /metrics")
	if err := srv.ListenAndServe(); err != nil && err != http.ErrServerClosed {
		return err
	}
	return nil
}

func metric(b *strings.Builder, name, kind, help string) {
	fmt.Fprintf(b, "# HELP %s %s\n# TYPE %s %s\n", name, help, name, kind)
}

func seconds(d time.Duration) string {
	return strconv.FormatFloat(d.Seconds(), 'f', 3, 64)
}

func unix(t time.Time) int64 {
	if t.IsZero() {
		return 0
	}
	return t.Unix()
}

func writeText(w http.ResponseWriter, code int, body string) {
	w.Header().Set("Content-Type", "text/plain; charset=utf-8")
	w.WriteHeader(code)
	_, _ = w.Write([]byte(body))
}