Dash/internal/checker/checker.go

// Package checker implements the four probe types: http, tcp, tls, icmp.
package checker

import (
	"context"
	"crypto/tls"
	"encoding/json"
	"fmt"
	"net"
	"net/http"
	"os"
	"strconv"
	"time"

	"golang.org/x/net/icmp"
	"golang.org/x/net/ipv4"

	"dash/internal/config"
)

// Result is the outcome of a single probe.
type Result struct {
	TS        time.Time
	OK        bool
	Degraded  bool    // ok but worrying (slow, cert nearing expiry)
	LatencyMS float64 // round-trip time
	Detail    string  // machine-readable extra: http status, cert days left
	Err       string
}

func fail(start time.Time, format string, args ...any) Result {
	return Result{
		TS:        start,
		LatencyMS: float64(time.Since(start).Microseconds()) / 1000.0,
		Err:       fmt.Sprintf(format, args...),
	}
}

// Run executes the probe described by t once.
func Run(ctx context.Context, t config.Target) Result {
	ctx, cancel := context.WithTimeout(ctx, t.Timeout.Duration)
	defer cancel()

	var res Result
	switch t.Type {
	case config.CheckHTTP:
		res = checkHTTP(ctx, t)
	case config.CheckTCP:
		res = checkTCP(ctx, t)
	case config.CheckTLS:
		res = checkTLS(ctx, t)
	case config.CheckICMP:
		res = checkICMP(ctx, t)
	case config.CheckDocker:
		res = checkDocker(ctx, t)
	default:
		res = fail(time.Now(), "unknown check type %q", t.Type)
	}
	if res.OK && t.DegradedLatencyMS > 0 && res.LatencyMS > t.DegradedLatencyMS {
		res.Degraded = true
	}
	return res
}

func checkHTTP(ctx context.Context, t config.Target) Result {
	start := time.Now()
	req, err := http.NewRequestWithContext(ctx, http.MethodGet, t.URL, nil)
	if err != nil {
		return fail(start, "bad url: %v", err)
	}
	req.Header.Set("User-Agent", "dash-monitor/1.0")
	client := &http.Client{
		// Each probe measures a full fresh connection, not a pooled one.
		Transport: &http.Transport{DisableKeepAlives: true},
		CheckRedirect: func(req *http.Request, via []*http.Request) error {
			return http.ErrUseLastResponse
		},
	}
	resp, err := client.Do(req)
	if err != nil {
		return fail(start, "%v", err)
	}
	defer resp.Body.Close()
	lat := float64(time.Since(start).Microseconds()) / 1000.0

	ok := resp.StatusCode < 400
	if t.ExpectStatus != 0 {
		ok = resp.StatusCode == t.ExpectStatus
	}
	res := Result{TS: start, OK: ok, LatencyMS: lat, Detail: strconv.Itoa(resp.StatusCode)}
	if !ok {
		res.Err = fmt.Sprintf("unexpected status %d", resp.StatusCode)
	}
	return res
}

func checkTCP(ctx context.Context, t config.Target) Result {
	start := time.Now()
	var d net.Dialer
	conn, err := d.DialContext(ctx, "tcp", t.Address)
	if err != nil {
		return fail(start, "%v", err)
	}
	conn.Close()
	return Result{TS: start, OK: true, LatencyMS: float64(time.Since(start).Microseconds()) / 1000.0}
}

func checkTLS(ctx context.Context, t config.Target) Result {
	start := time.Now()
	d := tls.Dialer{Config: &tls.Config{}}
	conn, err := d.DialContext(ctx, "tcp", t.Address)
	if err != nil {
		return fail(start, "%v", err)
	}
	defer conn.Close()
	lat := float64(time.Since(start).Microseconds()) / 1000.0

	state := conn.(*tls.Conn).ConnectionState()
	if len(state.PeerCertificates) == 0 {
		return fail(start, "no peer certificates")
	}
	leaf := state.PeerCertificates[0]
	days := int(time.Until(leaf.NotAfter).Hours() / 24)
	res := Result{TS: start, OK: true, LatencyMS: lat, Detail: strconv.Itoa(days)}
	switch {
	case days < 0:
		res.OK = false
		res.Err = fmt.Sprintf("certificate expired %d days ago", -days)
	case days < t.WarnDays:
		res.Degraded = true
		res.Err = fmt.Sprintf("certificate expires in %d days", days)
	}
	return res
}

// dockerInspect is the subset of the container-inspect response we read.
type dockerInspect struct {
	State struct {
		Status  string `json:"Status"`
		Running bool   `json:"Running"`
		Health  *struct {
			Status string `json:"Status"` // starting | healthy | unhealthy
		} `json:"Health"`
	} `json:"State"`
}

func checkDocker(ctx context.Context, t config.Target) Result {
	start := time.Now()
	client := &http.Client{
		Transport: &http.Transport{
			DialContext: func(ctx context.Context, _, _ string) (net.Conn, error) {
				var d net.Dialer
				return d.DialContext(ctx, "unix", t.Socket)
			},
		},
	}
	url := "http://docker/containers/" + t.Container + "/json"
	req, err := http.NewRequestWithContext(ctx, http.MethodGet, url, nil)
	if err != nil {
		return fail(start, "bad request: %v", err)
	}
	resp, err := client.Do(req)
	if err != nil {
		return fail(start, "docker socket: %v", err)
	}
	defer resp.Body.Close()
	lat := float64(time.Since(start).Microseconds()) / 1000.0

	if resp.StatusCode == http.StatusNotFound {
		return fail(start, "container %q not found", t.Container)
	}
	if resp.StatusCode != http.StatusOK {
		return fail(start, "docker api status %d", resp.StatusCode)
	}
	var info dockerInspect
	if err := json.NewDecoder(resp.Body).Decode(&info); err != nil {
		return fail(start, "decode: %v", err)
	}

	res := Result{TS: start, LatencyMS: lat, Detail: info.State.Status}
	if !info.State.Running {
		res.Err = fmt.Sprintf("container %s", info.State.Status)
		return res
	}
	res.OK = true
	if h := info.State.Health; h != nil {
		res.Detail = h.Status
		switch h.Status {
		case "unhealthy":
			res.OK = false
			res.Err = "healthcheck failing"
		case "starting":
			res.Degraded = true
			res.Err = "healthcheck starting"
		}
	}
	return res
}

var icmpSeq uint32

func checkICMP(ctx context.Context, t config.Target) Result {
	start := time.Now()
	dst, err := net.ResolveIPAddr("ip4", t.Host)
	if err != nil {
		return fail(start, "resolve: %v", err)
	}

	// Prefer unprivileged datagram ICMP; fall back to a raw socket (needs
	// root or CAP_NET_RAW).
	var peer net.Addr = &net.UDPAddr{IP: dst.IP}
	conn, err := icmp.ListenPacket("udp4", "0.0.0.0")
	if err != nil {
		conn, err = icmp.ListenPacket("ip4:icmp", "0.0.0.0")
		if err != nil {
			return fail(start, "icmp socket: %v (needs CAP_NET_RAW or net.ipv4.ping_group_range)", err)
		}
		peer = dst
	}
	defer conn.Close()

	icmpSeq++
	msg := icmp.Message{
		Type: ipv4.ICMPTypeEcho,
		Body: &icmp.Echo{
			ID:   os.Getpid() & 0xffff,
			Seq:  int(icmpSeq & 0xffff),
			Data: []byte("dash-monitor-ping"),
		},
	}
	wire, err := msg.Marshal(nil)
	if err != nil {
		return fail(start, "marshal: %v", err)
	}

	deadline, _ := ctx.Deadline()
	conn.SetDeadline(deadline)

	sent := time.Now()
	if _, err := conn.WriteTo(wire, peer); err != nil {
		return fail(start, "send: %v", err)
	}

	buf := make([]byte, 1500)
	for {
		n, _, err := conn.ReadFrom(buf)
		if err != nil {
			return fail(start, "recv: %v", err)
		}
		reply, err := icmp.ParseMessage(1, buf[:n]) // 1 = ICMPv4 protocol number
		if err != nil {
			continue
		}
		if reply.Type == ipv4.ICMPTypeEchoReply {
			lat := float64(time.Since(sent).Microseconds()) / 1000.0
			return Result{TS: start, OK: true, LatencyMS: lat}
		}
	}
}