diff --git a/internal/locate/region_request.go b/internal/locate/region_request.go
index b8e2dede0..68a6a3dee 100644
--- a/internal/locate/region_request.go
+++ b/internal/locate/region_request.go
@@ -2125,7 +2125,7 @@ func (s *RegionRequestSender) onRegionError(
 		if s.replicaSelector != nil {
 			return s.replicaSelector.onServerIsBusy(bo, ctx, req, serverIsBusy)
 		}
-		logutil.Logger(bo.GetCtx()).Warn(
+		logutil.Logger(bo.GetCtx()).Debug(
 			"tikv reports `ServerIsBusy` retry later",
 			zap.String("reason", regionErr.GetServerIsBusy().GetReason()),
 			zap.Stringer("ctx", ctx),
@@ -2232,7 +2232,7 @@ func (s *RegionRequestSender) onRegionError(
 	// This error is specific to stale read and the target replica is randomly selected. If the request is sent
 	// to the leader, the data must be ready, so we don't backoff here.
 	if regionErr.GetDataIsNotReady() != nil {
-		logutil.Logger(bo.GetCtx()).Warn(
+		logutil.Logger(bo.GetCtx()).Debug(
 			"tikv reports `DataIsNotReady` retry later",
 			zap.Uint64("store-id", ctx.Store.storeID),
 			zap.Uint64("peer-id", regionErr.GetDataIsNotReady().GetPeerId()),