diff --git a/pkg/kubeserver/client/manager.go b/pkg/kubeserver/client/manager.go
index e608e19b5..19a81dd8f 100644
--- a/pkg/kubeserver/client/manager.go
+++ b/pkg/kubeserver/client/manager.go
@@ -101,6 +101,9 @@ func (m *ClustersManager) AddClient(dbCluster manager.ICluster) error {
 func (m *ClustersManager) IsClusterClientHealthy(dbCluster manager.ICluster) bool {
 	clusterId := dbCluster.GetId()
 	man := m.getManager(clusterId)
+	if man == nil {
+		return false
+	}
 	err := man.ClientV2.K8S().IsReachable()
 	if err != nil {
 		log.Warningf("check cluster %s is reachable error: %v", dbCluster.GetName(), err)
diff --git a/pkg/kubeserver/models/clusters.go b/pkg/kubeserver/models/clusters.go
index 5b255311f..c50381453 100644
--- a/pkg/kubeserver/models/clusters.go
+++ b/pkg/kubeserver/models/clusters.go
@@ -923,8 +923,17 @@ func (m *SClusterManager) ClusterHealthCheckTask(ctx context.Context, userCred m
 			}
 			continue
 		} else {
-			c.SetStatus(ctx, userCred, api.ClusterStatusLost, err.Error())
-			client.GetClustersManager().RemoveClient(c.GetId())
+			if c.GetName() == SystemClusterName {
+				log.Infof("cluster %q is unhealthy, perform force sync", c.GetName())
+				if _, err := c.PerformSync(ctx, userCred, nil, api.ClusterSyncInput{
+					Force: true,
+				}); err != nil {
+					log.Errorf("cluster %q perform force sync when health check error: %v", c.GetName(), err)
+				}
+			} else {
+				c.SetStatus(ctx, userCred, api.ClusterStatusLost, err.Error())
+				client.GetClustersManager().RemoveClient(c.GetId())
+			}
 		}
 	}
 }