get stable dns

common/common.go

@@ -2,6 +2,7 @@ package common
 
 import (
 	"runtime/debug"
+	"sort"
 
 	"github.com/memory-overflow/highly-balanced-scheduling-agent/common/config"
 )
@@ -12,3 +13,18 @@ func Recover() {
 		config.GetLogger().Sugar().Fatalf("panic: %v, stacktrace: %s", p, debug.Stack())
 	}
 }
+
+// SliceSame 对于两个列表的值是否一样
+func SliceSame(a, b []string) bool {
+	if len(a) != len(b) {
+		return false
+	}
+	sort.Strings(a)
+	sort.Strings(b)
+	for i := 0; i < len(a); i++ {
+		if a[i] != b[i] {
+			return false
+		}
+	}
+	return true
+}

service/init.go

@@ -49,6 +49,17 @@ func syncIps(ctx context.Context, service *k8sserviceInfo) {
 	if err != nil {
 		return
 	}
+	// 三次拉取到的都是同一个 ip 列表再继续，防止 pod 重启过程中dns ip 列表不稳定
+	for i := 0; i < 2; i++ {
+		time.Sleep(100 * time.Millisecond)
+		dupips, err := common.GetDns(ctx, service.k8sHost)
+		if err != nil {
+			return
+		}
+		if !common.SliceSame(ips, dupips) {
+			return
+		}
+	}
 	var newMap sync.Map
 	for _, ip := range ips {
 		if value, ok := service.lastConnections.Load(ip); ok {

service/proxy_service.go

@@ -50,6 +50,7 @@ func (service *k8sserviceInfo) getIp(ctx context.Context) (ip string, last *int3
 				atomic.AddInt32(maxLast, -1)
 				return ip, maxLast, nil
 			}
+			time.Sleep(2 * time.Second)
 		}
 	}
 }

service/serve_http.go

@@ -59,6 +59,7 @@ func handle(ctx context.Context, svc *k8sserviceInfo) {
 			}
 			go func() {
 				defer atomic.AddInt32(last, 1) // 对应 ip 剩余连接 +1
+				config.GetLogger().Sugar().Infof("selected ip %s, last conn: %d", ip, *last)
 				transport(fmt.Sprintf("http://%s:%d%s", ip, svc.k8sPort, svc.uri), job.rw, job.req)
 				job.done <- struct{}{}
 			}()