cmd/{k8s-operator,containerboot},kube: ensure egress ProxyGroup proxies don't terminate while cluster traffic is still routed to them (#14436)

cmd/{containerboot,k8s-operator},kube: add preshutdown hook for egress PG proxies

This change is part of work towards minimizing downtime during update
rollouts of egress ProxyGroup replicas.
This change:
- updates the containerboot health check logic to return Pod IP in headers,
if set
- always runs the health check for egress PG proxies
- updates ClusterIP Services created for PG egress endpoints to include
the health check endpoint
- implements preshutdown endpoint in proxies. The preshutdown endpoint
logic waits till, for all currently configured egress services, the ClusterIP
Service health check endpoint is no longer returned by the shutting-down Pod
(by looking at the new Pod IP header).
- ensures that kubelet is configured to call the preshutdown endpoint

This reduces the possibility that, as replicas are terminated during an update,
a replica gets terminated to which cluster traffic is still being routed via
the ClusterIP Service because kube proxy has not yet updated routig rules.
This is not a perfect check as in practice, it only checks that the kube
proxy on the node on which the proxy runs has updated rules. However, overall
this might be good enough.

The preshutdown logic is disabled if users have configured a custom health check
port via TS_LOCAL_ADDR_PORT env var. This change throws a warnign if so and in
future setting of that env var for operator proxies might be disallowed (as users
shouldn't need to configure this for a Pod directly).
This is backwards compatible with earlier proxy versions.

Updates tailscale/tailscale#14326


Signed-off-by: Irbe Krumina <irbe@tailscale.com>

cmd/k8s-operator/proxygroup.go

@@ -32,6 +32,7 @@ import (
 	"tailscale.com/ipn"
 	tsoperator "tailscale.com/k8s-operator"
 	tsapi "tailscale.com/k8s-operator/apis/v1alpha1"
+	"tailscale.com/kube/egressservices"
 	"tailscale.com/kube/kubetypes"
 	"tailscale.com/tailcfg"
 	"tailscale.com/tstime"
@@ -166,6 +167,7 @@ func (r *ProxyGroupReconciler) Reconcile(ctx context.Context, req reconcile.Requ
 			r.recorder.Eventf(pg, corev1.EventTypeWarning, reasonProxyGroupCreationFailed, err.Error())
 			return setStatusReady(pg, metav1.ConditionFalse, reasonProxyGroupCreationFailed, err.Error())
 		}
+		validateProxyClassForPG(logger, pg, proxyClass)
 		if !tsoperator.ProxyClassIsReady(proxyClass) {
 			message := fmt.Sprintf("the ProxyGroup's ProxyClass %s is not yet in a ready state, waiting...", proxyClassName)
 			logger.Info(message)
@@ -204,6 +206,31 @@ func (r *ProxyGroupReconciler) Reconcile(ctx context.Context, req reconcile.Requ
 	return setStatusReady(pg, metav1.ConditionTrue, reasonProxyGroupReady, reasonProxyGroupReady)
 }
 
+// validateProxyClassForPG applies custom validation logic for ProxyClass applied to ProxyGroup.
+func validateProxyClassForPG(logger *zap.SugaredLogger, pg *tsapi.ProxyGroup, pc *tsapi.ProxyClass) {
+	if pg.Spec.Type == tsapi.ProxyGroupTypeIngress {
+		return
+	}
+	// Our custom logic for ensuring minimum downtime ProxyGroup update rollouts relies on the local health check
+	// beig accessible on the replica Pod IP:9002. This address can also be modified by users, via
+	// TS_LOCAL_ADDR_PORT env var.
+	//
+	// Currently TS_LOCAL_ADDR_PORT controls Pod's health check and metrics address. _Probably_ there is no need for
+	// users to set this to a custom value. Users who want to consume metrics, should integrate with the metrics
+	// Service and/or ServiceMonitor, rather than Pods directly. The health check is likely not useful to integrate
+	// directly with for operator proxies (and we should aim for unified lifecycle logic in the operator, users
+	// shouldn't need to set their own).
+	//
+	// TODO(irbekrm): maybe disallow configuring this env var in future (in Tailscale 1.84 or later).
+	if hasLocalAddrPortSet(pc) {
+		msg := fmt.Sprintf("ProxyClass %s applied to an egress ProxyGroup has TS_LOCAL_ADDR_PORT env var set to a custom value."+
+			"This will disable the ProxyGroup graceful failover mechanism, so you might experience downtime when ProxyGroup pods are restarted."+
+			"In future we will remove the ability to set custom TS_LOCAL_ADDR_PORT for egress ProxyGroups."+
+			"Please raise an issue if you expect that this will cause issues for your workflow.", pc.Name)
+		logger.Warn(msg)
+	}
+}
+
 func (r *ProxyGroupReconciler) maybeProvision(ctx context.Context, pg *tsapi.ProxyGroup, proxyClass *tsapi.ProxyClass) error {
 	logger := r.logger(pg.Name)
 	r.mu.Lock()
@@ -253,10 +280,11 @@ func (r *ProxyGroupReconciler) maybeProvision(ctx context.Context, pg *tsapi.Pro
 		return fmt.Errorf("error provisioning RoleBinding: %w", err)
 	}
 	if pg.Spec.Type == tsapi.ProxyGroupTypeEgress {
-		cm := pgEgressCM(pg, r.tsNamespace)
+		cm, hp := pgEgressCM(pg, r.tsNamespace)
 		if _, err := createOrUpdate(ctx, r.Client, r.tsNamespace, cm, func(existing *corev1.ConfigMap) {
 			existing.ObjectMeta.Labels = cm.ObjectMeta.Labels
 			existing.ObjectMeta.OwnerReferences = cm.ObjectMeta.OwnerReferences
+			mak.Set(&existing.BinaryData, egressservices.KeyHEPPings, hp)
 		}); err != nil {
 			return fmt.Errorf("error provisioning egress ConfigMap %q: %w", cm.Name, err)
 		}
@@ -270,7 +298,7 @@ func (r *ProxyGroupReconciler) maybeProvision(ctx context.Context, pg *tsapi.Pro
 			return fmt.Errorf("error provisioning ingress ConfigMap %q: %w", cm.Name, err)
 		}
 	}
-	ss, err := pgStatefulSet(pg, r.tsNamespace, r.proxyImage, r.tsFirewallMode)
+	ss, err := pgStatefulSet(pg, r.tsNamespace, r.proxyImage, r.tsFirewallMode, proxyClass)
 	if err != nil {
 		return fmt.Errorf("error generating StatefulSet spec: %w", err)
 	}