Initial revisin of healthchecker, inspired by HAProxy

internal/prober/icmp.go

@@ -0,0 +1,118 @@
+package prober
+
+import (
+	"context"
+	"fmt"
+	"math/rand/v2"
+	"net"
+	"time"
+
+	"golang.org/x/net/icmp"
+	"golang.org/x/net/ipv4"
+	"golang.org/x/net/ipv6"
+
+	"git.ipng.ch/ipng/vpp-maglev/internal/health"
+)
+
+// ICMPProbe sends an ICMP echo request to cfg.Target inside the healthcheck
+// netns and waits for a matching reply.
+func ICMPProbe(ctx context.Context, cfg ProbeConfig) health.ProbeResult {
+	isV6 := cfg.Target.To4() == nil
+
+	var ok bool
+	err := inNetns(cfg.HealthCheckNetns, func() error {
+		var network string
+		var proto int
+		var msgType icmp.Type
+		if isV6 {
+			network = "ip6:ipv6-icmp"
+			proto = 58
+			msgType = ipv6.ICMPTypeEchoRequest
+		} else {
+			network = "ip4:icmp"
+			proto = 1
+			msgType = ipv4.ICMPTypeEcho
+		}
+
+		listenAddr := ""
+		if cfg.ProbeSrc != nil {
+			listenAddr = cfg.ProbeSrc.String()
+		}
+		pc, err := net.ListenPacket(network, listenAddr)
+		if err != nil {
+			return fmt.Errorf("listen icmp (%s): %w", network, err)
+		}
+		defer pc.Close()
+
+		id := rand.IntN(0xffff) + 1
+		seq := rand.IntN(0xffff) + 1
+
+		msg := icmp.Message{
+			Type: msgType,
+			Code: 0,
+			Body: &icmp.Echo{
+				ID:   id,
+				Seq:  seq,
+				Data: []byte("maglev-hc"),
+			},
+		}
+		b, err := msg.Marshal(nil)
+		if err != nil {
+			return fmt.Errorf("marshal icmp: %w", err)
+		}
+
+		dst := &net.IPAddr{IP: cfg.Target}
+		if _, err := pc.WriteTo(b, dst); err != nil {
+			return fmt.Errorf("write icmp: %w", err)
+		}
+
+		deadline := time.Now().Add(cfg.Timeout)
+		pc.SetDeadline(deadline) //nolint:errcheck
+
+		buf := make([]byte, 1500)
+		for time.Now().Before(deadline) {
+			select {
+			case <-ctx.Done():
+				return ctx.Err()
+			default:
+			}
+
+			n, _, err := pc.ReadFrom(buf)
+			if err != nil {
+				if isTimeout(err) {
+					return nil
+				}
+				return fmt.Errorf("read icmp: %w", err)
+			}
+
+			reply, err := icmp.ParseMessage(proto, buf[:n])
+			if err != nil {
+				continue
+			}
+			echo, ok2 := reply.Body.(*icmp.Echo)
+			if !ok2 {
+				continue
+			}
+			if echo.ID == id && echo.Seq == seq {
+				ok = true
+				return nil
+			}
+		}
+		return nil
+	})
+
+	if err != nil {
+		return health.ProbeResult{OK: false, Layer: health.LayerL7, Code: "L7TOUT", Detail: err.Error()}
+	}
+	if ok {
+		return health.ProbeResult{OK: true, Layer: health.LayerL7, Code: "L7OK"}
+	}
+	return health.ProbeResult{OK: false, Layer: health.LayerL7, Code: "L7TOUT", Detail: "no reply received"}
+}
+
+func isTimeout(err error) bool {
+	if netErr, ok := err.(net.Error); ok {
+		return netErr.Timeout()
+	}
+	return false
+}