tailscale/cmd/k8s-operator/ingress.go

// Copyright (c) Tailscale Inc & AUTHORS
// SPDX-License-Identifier: BSD-3-Clause

//go:build !plan9

package main

import (
	"context"
	"fmt"
	"slices"
	"strings"
	"sync"

	"github.com/pkg/errors"
	"go.uber.org/zap"
	corev1 "k8s.io/api/core/v1"
	networkingv1 "k8s.io/api/networking/v1"
	apierrors "k8s.io/apimachinery/pkg/api/errors"
	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
	"k8s.io/apimachinery/pkg/types"
	"k8s.io/client-go/tools/record"
	"sigs.k8s.io/controller-runtime/pkg/client"
	"sigs.k8s.io/controller-runtime/pkg/reconcile"
	"tailscale.com/ipn"
	"tailscale.com/types/opt"
	"tailscale.com/util/clientmetric"
	"tailscale.com/util/set"
)

const (
	tailscaleIngressClassName      = "tailscale"                                   // ingressClass.metadata.name for tailscale IngressClass resource
	tailscaleIngressControllerName = "tailscale.com/ts-ingress"                    // ingressClass.spec.controllerName for tailscale IngressClass resource
	ingressClassDefaultAnnotation  = "ingressclass.kubernetes.io/is-default-class" // we do not support this https://kubernetes.io/docs/concepts/services-networking/ingress/#default-ingress-class
)

type IngressReconciler struct {
	client.Client

	recorder record.EventRecorder
	ssr      *tailscaleSTSReconciler
	logger   *zap.SugaredLogger

	mu sync.Mutex // protects following

	// managedIngresses is a set of all ingress resources that we're currently
	// managing. This is only used for metrics.
	managedIngresses set.Slice[types.UID]
}

var (
	// gaugeIngressResources tracks the number of ingress resources that we're
	// currently managing.
	gaugeIngressResources = clientmetric.NewGauge("k8s_ingress_resources")
)

func (a *IngressReconciler) Reconcile(ctx context.Context, req reconcile.Request) (_ reconcile.Result, err error) {
	logger := a.logger.With("ingress-ns", req.Namespace, "ingress-name", req.Name)
	logger.Debugf("starting reconcile")
	defer logger.Debugf("reconcile finished")

	ing := new(networkingv1.Ingress)
	err = a.Get(ctx, req.NamespacedName, ing)
	if apierrors.IsNotFound(err) {
		// Request object not found, could have been deleted after reconcile request.
		logger.Debugf("ingress not found, assuming it was deleted")
		return reconcile.Result{}, nil
	} else if err != nil {
		return reconcile.Result{}, fmt.Errorf("failed to get ing: %w", err)
	}
	if !ing.DeletionTimestamp.IsZero() || !a.shouldExpose(ing) {
		logger.Debugf("ingress is being deleted or should not be exposed, cleaning up")
		return reconcile.Result{}, a.maybeCleanup(ctx, logger, ing)
	}

	return reconcile.Result{}, a.maybeProvision(ctx, logger, ing)
}

func (a *IngressReconciler) maybeCleanup(ctx context.Context, logger *zap.SugaredLogger, ing *networkingv1.Ingress) error {
	ix := slices.Index(ing.Finalizers, FinalizerName)
	if ix < 0 {
		logger.Debugf("no finalizer, nothing to do")
		a.mu.Lock()
		defer a.mu.Unlock()
		a.managedIngresses.Remove(ing.UID)
		gaugeIngressResources.Set(int64(a.managedIngresses.Len()))
		return nil
	}

	if done, err := a.ssr.Cleanup(ctx, logger, childResourceLabels(ing.Name, ing.Namespace, "ingress")); err != nil {
		return fmt.Errorf("failed to cleanup: %w", err)
	} else if !done {
		logger.Debugf("cleanup not done yet, waiting for next reconcile")
		return nil
	}

	ing.Finalizers = append(ing.Finalizers[:ix], ing.Finalizers[ix+1:]...)
	if err := a.Update(ctx, ing); err != nil {
		return fmt.Errorf("failed to remove finalizer: %w", err)
	}

	// Unlike most log entries in the reconcile loop, this will get printed
	// exactly once at the very end of cleanup, because the final step of
	// cleanup removes the tailscale finalizer, which will make all future
	// reconciles exit early.
	logger.Infof("unexposed ingress from tailnet")
	a.mu.Lock()
	defer a.mu.Unlock()
	a.managedIngresses.Remove(ing.UID)
	gaugeIngressResources.Set(int64(a.managedIngresses.Len()))
	return nil
}

// maybeProvision ensures that ing is exposed over tailscale, taking any actions
// necessary to reach that state.
//
// This function adds a finalizer to ing, ensuring that we can handle orderly
// deprovisioning later.
func (a *IngressReconciler) maybeProvision(ctx context.Context, logger *zap.SugaredLogger, ing *networkingv1.Ingress) error {
	if err := a.validateIngressClass(ctx); err != nil {
		logger.Warnf("error validating tailscale IngressClass: %v. In future this might be a terminal error.", err)

	}
	if !slices.Contains(ing.Finalizers, FinalizerName) {
		// This log line is printed exactly once during initial provisioning,
		// because once the finalizer is in place this block gets skipped. So,
		// this is a nice place to tell the operator that the high level,
		// multi-reconcile operation is underway.
		logger.Infof("exposing ingress over tailscale")
		ing.Finalizers = append(ing.Finalizers, FinalizerName)
		if err := a.Update(ctx, ing); err != nil {
			return fmt.Errorf("failed to add finalizer: %w", err)
		}
	}
	a.mu.Lock()
	a.managedIngresses.Add(ing.UID)
	gaugeIngressResources.Set(int64(a.managedIngresses.Len()))
	a.mu.Unlock()

	if !a.ssr.IsHTTPSEnabledOnTailnet() {
		a.recorder.Event(ing, corev1.EventTypeWarning, "HTTPSNotEnabled", "HTTPS is not enabled on the tailnet; ingress may not work")
	}

	// magic443 is a fake hostname that we can use to tell containerboot to swap
	// out with the real hostname once it's known.
	const magic443 = "${TS_CERT_DOMAIN}:443"
	sc := &ipn.ServeConfig{
		TCP: map[uint16]*ipn.TCPPortHandler{
			443: {
				HTTPS: true,
			},
		},
		Web: map[ipn.HostPort]*ipn.WebServerConfig{
			magic443: {
				Handlers: map[string]*ipn.HTTPHandler{},
			},
		},
	}
	if opt.Bool(ing.Annotations[AnnotationFunnel]).EqualBool(true) {
		sc.AllowFunnel = map[ipn.HostPort]bool{
			magic443: true,
		}
	}

	web := sc.Web[magic443]
	addIngressBackend := func(b *networkingv1.IngressBackend, path string) {
		if b == nil {
			return
		}
		if b.Service == nil {
			a.recorder.Eventf(ing, corev1.EventTypeWarning, "InvalidIngressBackend", "backend for path %q is missing service", path)
			return
		}
		var svc corev1.Service
		if err := a.Get(ctx, types.NamespacedName{Namespace: ing.Namespace, Name: b.Service.Name}, &svc); err != nil {
			a.recorder.Eventf(ing, corev1.EventTypeWarning, "InvalidIngressBackend", "failed to get service %q for path %q: %v", b.Service.Name, path, err)
			return
		}
		if svc.Spec.ClusterIP == "" || svc.Spec.ClusterIP == "None" {
			a.recorder.Eventf(ing, corev1.EventTypeWarning, "InvalidIngressBackend", "backend for path %q has invalid ClusterIP", path)
			return
		}
		var port int32
		if b.Service.Port.Name != "" {
			for _, p := range svc.Spec.Ports {
				if p.Name == b.Service.Port.Name {
					port = p.Port
					break
				}
			}
		} else {
			port = b.Service.Port.Number
		}
		if port == 0 {
			a.recorder.Eventf(ing, corev1.EventTypeWarning, "InvalidIngressBackend", "backend for path %q has invalid port", path)
			return
		}
		proto := "http://"
		if port == 443 || b.Service.Port.Name == "https" {
			proto = "https+insecure://"
		}
		web.Handlers[path] = &ipn.HTTPHandler{
			Proxy: proto + svc.Spec.ClusterIP + ":" + fmt.Sprint(port) + path,
		}
	}
	addIngressBackend(ing.Spec.DefaultBackend, "/")

	var tlsHost string // hostname or FQDN or empty
	if ing.Spec.TLS != nil && len(ing.Spec.TLS) > 0 && len(ing.Spec.TLS[0].Hosts) > 0 {
		tlsHost = ing.Spec.TLS[0].Hosts[0]
	}
	for _, rule := range ing.Spec.Rules {
		// Host is optional, but if it's present it must match the TLS host
		// otherwise we ignore the rule.
		if rule.Host != "" && rule.Host != tlsHost {
			a.recorder.Eventf(ing, corev1.EventTypeWarning, "InvalidIngressBackend", "rule with host %q ignored, unsupported", rule.Host)
			continue
		}
		for _, p := range rule.HTTP.Paths {
			// Send a warning if folks use Exact path type - to make
			// it easier for us to support Exact path type matching
			// in the future if needed.
			// https://kubernetes.io/docs/concepts/services-networking/ingress/#path-types
			if *p.PathType == networkingv1.PathTypeExact {
				msg := "Exact path type strict matching is currently not supported and requests will be routed as for Prefix path type. This behaviour might change in the future."
				logger.Warnf(fmt.Sprintf("Unsupported Path type exact for path %s. %s", p.Path, msg))
				a.recorder.Eventf(ing, corev1.EventTypeWarning, "UnsupportedPathTypeExact", msg)
			}
			addIngressBackend(&p.Backend, p.Path)
		}
	}

	if len(web.Handlers) == 0 {
		logger.Warn("Ingress contains no valid backends")
		a.recorder.Eventf(ing, corev1.EventTypeWarning, "NoValidBackends", "no valid backends")
		return nil
	}

	crl := childResourceLabels(ing.Name, ing.Namespace, "ingress")
	var tags []string
	if tstr, ok := ing.Annotations[AnnotationTags]; ok {
		tags = strings.Split(tstr, ",")
	}
	hostname := ing.Namespace + "-" + ing.Name + "-ingress"
	if tlsHost != "" {
		hostname, _, _ = strings.Cut(tlsHost, ".")
	}

	sts := &tailscaleSTSConfig{
		Hostname:            hostname,
		ParentResourceName:  ing.Name,
		ParentResourceUID:   string(ing.UID),
		ServeConfig:         sc,
		Tags:                tags,
		ChildResourceLabels: crl,
	}

	if val := ing.GetAnnotations()[AnnotationExperimentalForwardClusterTrafficViaL7IngresProxy]; val == "true" {
		sts.ForwardClusterTrafficViaL7IngressProxy = true
	}

	if _, err := a.ssr.Provision(ctx, logger, sts); err != nil {
		return fmt.Errorf("failed to provision: %w", err)
	}

	_, tsHost, _, err := a.ssr.DeviceInfo(ctx, crl)
	if err != nil {
		return fmt.Errorf("failed to get device ID: %w", err)
	}
	if tsHost == "" {
		logger.Debugf("no Tailscale hostname known yet, waiting for proxy pod to finish auth")
		// No hostname yet. Wait for the proxy pod to auth.
		ing.Status.LoadBalancer.Ingress = nil
		if err := a.Status().Update(ctx, ing); err != nil {
			return fmt.Errorf("failed to update ingress status: %w", err)
		}
		return nil
	}

	logger.Debugf("setting ingress hostname to %q", tsHost)
	ing.Status.LoadBalancer.Ingress = []networkingv1.IngressLoadBalancerIngress{
		{
			Hostname: tsHost,
			Ports: []networkingv1.IngressPortStatus{
				{
					Protocol: "TCP",
					Port:     443,
				},
			},
		},
	}
	if err := a.Status().Update(ctx, ing); err != nil {
		return fmt.Errorf("failed to update ingress status: %w", err)
	}
	return nil
}

func (a *IngressReconciler) shouldExpose(ing *networkingv1.Ingress) bool {
	return ing != nil &&
		ing.Spec.IngressClassName != nil &&
		*ing.Spec.IngressClassName == tailscaleIngressClassName
}

// validateIngressClass attempts to validate that 'tailscale' IngressClass
// included in Tailscale installation manifests exists and has not been modified
// to attempt to enable features that we do not support.
func (a *IngressReconciler) validateIngressClass(ctx context.Context) error {
	ic := &networkingv1.IngressClass{
		ObjectMeta: metav1.ObjectMeta{
			Name: tailscaleIngressClassName,
		},
	}
	if err := a.Get(ctx, client.ObjectKeyFromObject(ic), ic); apierrors.IsNotFound(err) {
		return errors.New("Tailscale IngressClass not found in cluster. Latest installation manifests include a tailscale IngressClass - please update")
	} else if err != nil {
		return fmt.Errorf("error retrieving 'tailscale' IngressClass: %w", err)
	}
	if ic.Spec.Controller != tailscaleIngressControllerName {
		return fmt.Errorf("Tailscale Ingress class controller name %s does not match tailscale Ingress controller name %s. Ensure that you are using 'tailscale' IngressClass from latest Tailscale installation manifests", ic.Spec.Controller, tailscaleIngressControllerName)
	}
	if ic.GetAnnotations()[ingressClassDefaultAnnotation] != "" {
		return fmt.Errorf("%s annotation is set on 'tailscale' IngressClass, but Tailscale Ingress controller does not support default Ingress class. Ensure that you are using 'tailscale' IngressClass from latest Tailscale installation manifests", ingressClassDefaultAnnotation)
	}
	return nil
}