diff --git a/src/graph/service/QueryEngine.cpp b/src/graph/service/QueryEngine.cpp
index 6c9bb2b2b35..674d5bb0329 100644
--- a/src/graph/service/QueryEngine.cpp
+++ b/src/graph/service/QueryEngine.cpp
@@ -41,7 +41,7 @@ Status QueryEngine::init(std::shared_ptr<folly::IOThreadPoolExecutor> ioExecutor
   }
   optimizer_ = std::make_unique<opt::Optimizer>(rulesets);
 
-  return setupBackgroundThread();
+  return setupMemoryMonitorThread();
 }
 
 void QueryEngine::execute(RequestContextPtr rctx) {
@@ -55,9 +55,9 @@ void QueryEngine::execute(RequestContextPtr rctx) {
   instance->execute();
 }
 
-Status QueryEngine::setupBackgroundThread() {
-  bgThread_ = std::make_unique<thread::GenericWorker>();
-  if (!bgThread_ || !bgThread_->start("query-engine-bg")) {
+Status QueryEngine::setupMemoryMonitorThread() {
+  memoryMonitorThread_ = std::make_unique<thread::GenericWorker>();
+  if (!memoryMonitorThread_ || !memoryMonitorThread_->start("query-engine-bg")) {
     return Status::Error("Fail to start query engine background thread.");
   }
 
@@ -71,12 +71,7 @@ Status QueryEngine::setupBackgroundThread() {
   // Just to test whether to get the right memory info
   NG_RETURN_IF_ERROR(updateMemoryWatermark());
 
-  bgThread_->addRepeatTask(FLAGS_check_memory_interval_in_secs, [updateMemoryWatermark]() {
-    auto status = updateMemoryWatermark();
-    if (!status.ok()) {
-      LOG(ERROR) << status;
-    }
-  });
+  memoryMonitorThread_->addRepeatTask(FLAGS_check_memory_interval_in_secs, updateMemoryWatermark
 
   return Status::OK();
 }
diff --git a/src/graph/service/QueryEngine.h b/src/graph/service/QueryEngine.h
index 39e5e342f84..66efd23cb5f 100644
--- a/src/graph/service/QueryEngine.h
+++ b/src/graph/service/QueryEngine.h
@@ -20,15 +20,14 @@
 #include "graph/service/RequestContext.h"
 #include "interface/gen-cpp2/GraphService.h"
 
+namespace nebula {
+namespace graph {
+
 /**
  * QueryEngine is responsible to create and manage ExecutionPlan.
  * For the time being, we don't have the execution plan cache support,
  * instead we create a plan for each query, and destroy it upon finish.
  */
-
-namespace nebula {
-namespace graph {
-
 class QueryEngine final : public cpp::NonCopyable, public cpp::NonMovable {
  public:
   QueryEngine() = default;
@@ -43,13 +42,13 @@ class QueryEngine final : public cpp::NonCopyable, public cpp::NonMovable {
   const meta::MetaClient* metaClient() const { return metaClient_; }
 
  private:
-  Status setupBackgroundThread();
+  Status setupMemoryMonitorThread();
 
   std::unique_ptr<meta::SchemaManager> schemaManager_;
   std::unique_ptr<meta::IndexManager> indexManager_;
   std::unique_ptr<storage::GraphStorageClient> storage_;
   std::unique_ptr<opt::Optimizer> optimizer_;
-  std::unique_ptr<thread::GenericWorker> bgThread_;
+  std::unique_ptr<thread::GenericWorker> memoryMonitorThread_;
   meta::MetaClient* metaClient_{nullptr};
   CharsetInfo* charsetInfo_{nullptr};
 };