From be7fd17412bfb756208df00c8fa1df0082d69904 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E5=BC=A0=E7=A5=96=E5=BB=BA?= <zhangzujian.7@gmail.com>
Date: Sat, 20 Jul 2024 10:46:08 +0800
Subject: [PATCH] fix dialing https server (#4324)

Signed-off-by: zhangzujian <zhangzujian.7@gmail.com>
---
 .../kube-ovn/templates/controller-deploy.yaml |  2 +
 .../controller_health_check.go                | 38 +++++++++++++++----
 dist/images/install.sh                        |  2 +
 pkg/util/k8s.go                               | 37 ++++++++++++++----
 4 files changed, 65 insertions(+), 14 deletions(-)

diff --git a/charts/kube-ovn/templates/controller-deploy.yaml b/charts/kube-ovn/templates/controller-deploy.yaml
index 13cc27976d0..5c43dd6a178 100644
--- a/charts/kube-ovn/templates/controller-deploy.yaml
+++ b/charts/kube-ovn/templates/controller-deploy.yaml
@@ -169,12 +169,14 @@ spec:
             exec:
               command:
                 - /kube-ovn/kube-ovn-controller-healthcheck
+                - --tls={{- .Values.func.SECURE_SERVING }}
             periodSeconds: 3
             timeoutSeconds: 45
           livenessProbe:
             exec:
               command:
                 - /kube-ovn/kube-ovn-controller-healthcheck
+                - --tls={{- .Values.func.SECURE_SERVING }}
             initialDelaySeconds: 300
             periodSeconds: 7
             failureThreshold: 5
diff --git a/cmd/controller_health_check/controller_health_check.go b/cmd/controller_health_check/controller_health_check.go
index d77fbe83934..7e04769210f 100644
--- a/cmd/controller_health_check/controller_health_check.go
+++ b/cmd/controller_health_check/controller_health_check.go
@@ -1,25 +1,49 @@
 package controller_health_check
 
 import (
-	"net"
+	"flag"
 	"os"
 	"time"
 
+	"github.com/spf13/pflag"
+	"k8s.io/klog/v2"
+
 	"github.com/kubeovn/kube-ovn/pkg/util"
 )
 
 func CmdMain() {
+	tls := pflag.Bool("tls", false, "Whether kube-ovn-controller uses TLS")
+
+	klogFlags := flag.NewFlagSet("klog", flag.ExitOnError)
+	klog.InitFlags(klogFlags)
+
+	// sync the glog and klog flags.
+	pflag.CommandLine.VisitAll(func(f1 *pflag.Flag) {
+		f2 := klogFlags.Lookup(f1.Name)
+		if f2 != nil {
+			value := f1.Value.String()
+			if err := f2.Value.Set(value); err != nil {
+				util.LogFatalAndExit(err, "failed to set pflag")
+			}
+		}
+	})
+
+	pflag.CommandLine.AddGoFlagSet(klogFlags)
+	pflag.CommandLine.AddGoFlagSet(flag.CommandLine)
+	pflag.Parse()
+
 	addr := "127.0.0.1:10660"
 	if os.Getenv("ENABLE_BIND_LOCAL_IP") == "true" {
 		addr = util.JoinHostPort(os.Getenv("POD_IP"), 10660)
 	}
 
-	conn, err := net.DialTimeout("tcp", addr, 3*time.Second)
-	if err != nil {
-		util.LogFatalAndExit(err, "failed to probe the socket")
+	if *tls {
+		addr = "tls://" + addr
+	} else {
+		addr = "tcp://" + addr
 	}
-	err = conn.Close()
-	if err != nil {
-		util.LogFatalAndExit(err, "failed to close connection")
+
+	if err := util.DialTCP(addr, time.Second, false); err != nil {
+		util.LogFatalAndExit(err, "failed to probe the socket")
 	}
 }
diff --git a/dist/images/install.sh b/dist/images/install.sh
index de8459e5b2f..6dc342a0d1a 100755
--- a/dist/images/install.sh
+++ b/dist/images/install.sh
@@ -4209,12 +4209,14 @@ spec:
             exec:
               command:
                 - /kube-ovn/kube-ovn-controller-healthcheck
+                - --tls=${SECURE_SERVING}
             periodSeconds: 3
             timeoutSeconds: 45
           livenessProbe:
             exec:
               command:
                 - /kube-ovn/kube-ovn-controller-healthcheck
+                - --tls=${SECURE_SERVING}
             initialDelaySeconds: 300
             periodSeconds: 7
             failureThreshold: 5
diff --git a/pkg/util/k8s.go b/pkg/util/k8s.go
index 1ade1cc1bdc..bec7946c4bb 100644
--- a/pkg/util/k8s.go
+++ b/pkg/util/k8s.go
@@ -2,6 +2,7 @@ package util
 
 import (
 	"context"
+	"crypto/tls"
 	"encoding/json"
 	"fmt"
 	"net"
@@ -18,24 +19,46 @@ import (
 	"k8s.io/klog/v2"
 )
 
-func DialAPIServer(host string) error {
+func DialTCP(host string, timeout time.Duration, verbose bool) error {
 	u, err := url.Parse(host)
 	if err != nil {
 		return fmt.Errorf("failed to parse host %q: %v", host, err)
 	}
 
+	var conn net.Conn
 	address := net.JoinHostPort(u.Hostname(), u.Port())
-	timer := time.NewTimer(3 * time.Second)
+	switch u.Scheme {
+	case "tcp", "http":
+		conn, err = net.DialTimeout("tcp", address, timeout)
+	case "tls", "https":
+		config := &tls.Config{InsecureSkipVerify: true} // #nosec G402
+		conn, err = tls.DialWithDialer(&net.Dialer{Timeout: timeout}, "tcp", address, config)
+	default:
+		return fmt.Errorf("unsupported scheme %q", u.Scheme)
+	}
+
+	if err == nil {
+		if verbose {
+			klog.Infof("succeeded to dial host %q", host)
+		}
+		_ = conn.Close()
+		return nil
+	}
+
+	return fmt.Errorf("timed out dialing host %q", host)
+}
+
+func DialAPIServer(host string) error {
+	interval := 3 * time.Second
+	timer := time.NewTimer(interval)
 	for i := 0; i < 10; i++ {
-		conn, err := net.DialTimeout("tcp", address, 3*time.Second)
+		err := DialTCP(host, interval, true)
 		if err == nil {
-			klog.Infof("succeeded to dial apiserver %q", address)
-			_ = conn.Close()
 			return nil
 		}
-		klog.Warningf("failed to dial apiserver %q: %v", address, err)
+		klog.Warningf("failed to dial apiserver %q: %v", host, err)
 		<-timer.C
-		timer.Reset(3 * time.Second)
+		timer.Reset(interval)
 	}
 
 	return fmt.Errorf("timed out dialing apiserver %q", host)