feat(cluster): actively switch leader to updated pod.

mysqlcluster/syncer/statefulset.go

@@ -328,10 +328,14 @@ func (s *StatefulSetSyncer) updatePod(ctx context.Context) error {
 		return err
 	}
 	var leaderPod corev1.Pod
+	// newLeader saves the nodes (follower) that have been updated. 
+	// The old leader will switch the leader to the newLeader after all follower nodes are updated.
+	newLeader := ""
 	for _, pod := range pods.Items {
 		// Check if the pod is healthy.
 		if pod.ObjectMeta.Labels["healthy"] != "yes" {
-			return fmt.Errorf("can't start/continue 'update': pod[%s] is unhealthy", pod.Name)
+			s.log.V(1).Info("can't start/continue 'update': pod is unhealthy", "pod", pod.Name)
+			continue
 		}
 		// Skip if pod is leader.
 		if pod.ObjectMeta.Labels["role"] == string(utils.Leader) && leaderPod.Name == "" {
@@ -342,9 +346,14 @@ func (s *StatefulSetSyncer) updatePod(ctx context.Context) error {
 		if err := s.applyNWait(ctx, &pod); err != nil {
 			return err
 		}
+		newLeader = fmt.Sprintf("%s.%s.%s", pod.Name, s.GetNameForResource(utils.HeadlessSVC), pod.Namespace)
 	}
 	// There may be a case where Leader does not exist during the update process.
-	if leaderPod.Name != "" {
+	if leaderPod.Name != "" && newLeader != "" {
+		if err := s.XenonExecutor.RaftTryToLeader(newLeader); err != nil {
+			return err
+		}
+		s.log.V(1).Info("leader switch to", "pod", newLeader)
 		// Update the leader.
 		if err := s.applyNWait(ctx, &leaderPod); err != nil {
 			return err