diff --git a/server/coordinator.go b/server/coordinator.go
index b27c4af1b86..f800d949f55 100644
--- a/server/coordinator.go
+++ b/server/coordinator.go
@@ -206,6 +206,9 @@ func (c *coordinator) run() {
 	scheduleCfg := c.cluster.opt.load()
 	for _, schedulerCfg := range scheduleCfg.Schedulers {
 		if schedulerCfg.Disable {
+			scheduleCfg.Schedulers[k] = schedulerCfg
+			k++
+			log.Info("skip create ", schedulerCfg.Type)
 			continue
 		}
 		s, err := schedule.CreateScheduler(schedulerCfg.Type, c.limiter, schedulerCfg.Args...)
diff --git a/server/coordinator_test.go b/server/coordinator_test.go
index 52e0c8b9e8f..2eda488d67d 100644
--- a/server/coordinator_test.go
+++ b/server/coordinator_test.go
@@ -479,6 +479,14 @@ func (s *testCoordinatorSuite) TestPersistScheduler(c *C) {
 	c.Assert(newOpt.GetSchedulers(), HasLen, 7)
 	tc.clusterInfo.opt = newOpt
 
+	co = newCoordinator(tc.clusterInfo, hbStreams, namespace.DefaultClassifier)
+	co.run()
+	c.Assert(co.schedulers, HasLen, 3)
+	co.stop()
+	// suppose restart PD again
+	_, newOpt = newTestScheduleConfig()
+	newOpt.reload(tc.kv)
+	tc.clusterInfo.opt = newOpt
 	co = newCoordinator(tc.clusterInfo, hbStreams, namespace.DefaultClassifier)
 	co.run()
 	c.Assert(co.schedulers, HasLen, 3)