cert-status-exporter/main.go

package main

import (
	"crypto/tls"
	"errors"
	"fmt"
	"net/http"
	"os"
	"strconv"
	"strings"
	"time"

	"github.com/go-co-op/gocron/v2"
	"github.com/prometheus/client_golang/prometheus"
	"github.com/prometheus/client_golang/prometheus/promhttp"

	"github.com/urfave/cli/v2"

	log "github.com/sirupsen/logrus"
)

type CertificateMetrics struct {
	Domain string
	Port   int
	ALPN   []string

	ExpiryIn prometheus.Gauge
	IsValid  prometheus.Gauge
}

func NewCertificateMetrics(domain string, port int, alpn []string, name string) CertificateMetrics {
	metrics := CertificateMetrics{
		Domain: domain,
		Port:   port,
		ALPN:   alpn,
		ExpiryIn: prometheus.NewGauge(
			prometheus.GaugeOpts{
				Name:        "cert_status_expiry_in",
				ConstLabels: prometheus.Labels{"name": name},
			},
		),
		IsValid: prometheus.NewGauge(
			prometheus.GaugeOpts{
				Name:        "cert_status_is_valid",
				ConstLabels: prometheus.Labels{"name": name},
			},
		),
	}

	// Default values
	metrics.ExpiryIn.Set(0)
	metrics.IsValid.Set(0)
	return metrics
}

func (c *CertificateMetrics) register(registry *prometheus.Registry) {
	registry.MustRegister(
		c.ExpiryIn,
		c.IsValid,
	)
}

func (c *CertificateMetrics) checkTls() {
	conn, err := tls.Dial("tcp", c.Domain+":"+fmt.Sprint(c.Port), &tls.Config{
		NextProtos: c.ALPN,
	})
	if err != nil {
		log.Debugf("Failed to dial %s:%d@%s using ALPN %v: %v", c.Domain, c.Port, "tcp", c.ALPN, err)
		c.IsValid.Set(0)
		return
	}

	now := time.Now()
	leafCert := conn.ConnectionState().PeerCertificates[0]
	invalidIn := leafCert.NotAfter.Sub(now)
	c.ExpiryIn.Set(invalidIn.Hours() / 24)

	if leafCert.NotAfter.Compare(now) == -1 {
		log.Debugf("leafCert.NotAfter (%v) < now (%v)!!", leafCert.NotAfter, now)
		c.IsValid.Set(0)
	} else {
		if err := leafCert.VerifyHostname(c.Domain); err != nil {
			log.Debugf("Peer cert does not verify domain '%s'", c.Domain)
			c.IsValid.Set(0)
		} else {
			c.IsValid.Set(1)
		}
	}
}

func updateMetrics(metrics *[]CertificateMetrics, lastUpdated prometheus.Gauge) {
	log.Debugf("Updating metrics for %d domains...", len(*metrics))
	lastUpdated.SetToCurrentTime()
	for _, metric := range *metrics {
		metric.checkTls()
	}
	log.Debugf("Update done")
}

func run(ctx *cli.Context) error {
	debug := ctx.Bool("debug")
	host := ctx.String("host")
	port := ctx.Int("port")
	domains := ctx.StringSlice("domain")

	if debug {
		log.SetLevel(log.DebugLevel)
	}

	// Setup metrics
	registry := prometheus.NewRegistry()
	lastUpdatedMetric := prometheus.NewGauge(
		prometheus.GaugeOpts{
			Name: "certs_status_last_updated",
		},
	)
	registry.MustRegister(lastUpdatedMetric)

	metrics := make([]CertificateMetrics, 0)
	for _, d := range domains {
		log.Debugf("Parsing '%s'...", d)
		parts := strings.Split(d, ":")
		if len(parts) < 3 {
			log.Errorf("Invalid domain format for '%s'", d)
			return errors.New("Invalid domain format: Expects <domain>:<port>:<alpn>")
		}

		name := ""
		if len(parts) == 4 {
			name = parts[3]
		} else {
			name = parts[0]
		}

		port, err := strconv.Atoi(parts[1])
		if err != nil {
			log.Errorf("Failed to parse port of '%s'", d)
			return err
		}

		// Create the metric, and register it
		// TODO: Make this prettier
		alpn := strings.Split(parts[2], ";")
		log.Debugf("Registering: domain='%s' port='%d' alpn=%v name='%s'", parts[0], port, alpn, name)
		metric := NewCertificateMetrics(
			parts[0],
			port,
			alpn,
			name,
		)
		metric.register(registry)
		metrics = append(metrics, metric)
	}
	log.Infof("Parsed %d domain(s)", len(domains))

	// Setup the task scheduler
	scheduler, err := gocron.NewScheduler()
	if err != nil {
		log.Error("Failed to set up task scheduler")
		return err
	}
	_, err = scheduler.NewJob(
		gocron.DurationJob(24*time.Hour),
		gocron.NewTask(
			func() {
				updateMetrics(&metrics, lastUpdatedMetric)
			},
		),
	)
	if err != nil {
		log.Error("Failed to create periodic task")
		return err
	}

	// Perform an initial run to populate the metrics
	log.Info("Performing initial requests...")
	updateMetrics(&metrics, lastUpdatedMetric)
	log.Debug("Done")

	http.Handle(
		"/metrics", promhttp.HandlerFor(
			registry,
			promhttp.HandlerOpts{
				EnableOpenMetrics: true,
			},
		),
	)
	addr := host + ":" + fmt.Sprint(port)
	log.Infof("Handling requests at %s", addr)
	return http.ListenAndServe(addr, nil)
}

func main() {
	app := &cli.App{
		Action: run,
		Flags: []cli.Flag{
			&cli.BoolFlag{
				Name:  "debug",
				Value: false,
			},
			&cli.StringFlag{
				Name:  "host",
				Usage: "Host to expose metrics on",
				Value: "0.0.0.0",
			},
			&cli.IntFlag{
				Name:  "port",
				Usage: "Port to expose metrics on",
				Value: 8888,
			},
			&cli.StringSliceFlag{
				Name:     "domain",
				Usage:    "<domain>:<port>",
				Required: true,
			},
		},
	}
	if err := app.Run(os.Args); err != nil {
		log.Fatal(err)
	}
}