diff --git a/dbm-ui/backend/flow/engine/bamboo/scene/redis/redis_backend_scale.py b/dbm-ui/backend/flow/engine/bamboo/scene/redis/redis_backend_scale.py
index 41e8879ef1..cc08612380 100644
--- a/dbm-ui/backend/flow/engine/bamboo/scene/redis/redis_backend_scale.py
+++ b/dbm-ui/backend/flow/engine/bamboo/scene/redis/redis_backend_scale.py
@@ -10,6 +10,7 @@
 """
 import logging.config
 from collections import defaultdict
+from copy import deepcopy
 from dataclasses import asdict
 from typing import Dict, Optional
 
@@ -98,6 +99,9 @@ def __pre_check(
         else:
             if len(master_ips) != group_num:
                 raise Exception("new machine num != group_num.")
+        # shard_num 必须大于 group_num
+        if new_shard_num < group_num:
+            raise Exception("shard_num:{} < group_num:{}".format(new_shard_num, group_num))
         if old_shard_num != new_shard_num:
             raise Exception("old_shard_num {} != new_shard_num {}.".format(old_shard_num, new_shard_num))
         # 2024-05 要求可以不用整除，允许机器上的实例有多有少
@@ -401,6 +405,42 @@ def get_redis_install_sub_pipelines(
             redis_install_sub_pipelines.append(RedisBatchInstallAtomJob(self.root_id, self.data, act_kwargs, params))
         return redis_install_sub_pipelines
 
+    # > 4.0 版本需要重做一下slave，否则可能会丢数据
+    # rediscluster架构需要特殊考虑
+    def redis_local_redo_dr(self, act_kwargs, sync_relations) -> list:
+        sub_pipelines, resync_args = [], deepcopy(act_kwargs)
+        resync_args.cluster = {
+            "bk_biz_id": int(act_kwargs.cluster["bk_biz_id"]),
+            "cluster_id": int(act_kwargs.cluster["cluster_id"]),
+            "cluster_type": act_kwargs.cluster["cluster_type"],
+            "immute_domain": act_kwargs.cluster["immute_domain"],
+            "bk_cloud_id": int(act_kwargs.cluster["bk_cloud_id"]),
+            "instances": [],
+        }
+        for sync_link in sync_relations:
+            one_resync_args = deepcopy(resync_args)
+            new_master, new_slave = sync_link["sync_dst1"], sync_link["sync_dst2"]
+            for instances in sync_link["ins_link"]:
+                master_port, slave_port = instances["sync_dst1"], instances["sync_dst2"]
+                one_resync_args.cluster["instances"].append(
+                    {
+                        "master_ip": new_master,
+                        "master_port": int(master_port),
+                        "slave_ip": new_slave,
+                        "slave_port": int(slave_port),
+                    }
+                )
+            one_resync_args.exec_ip = new_slave
+            one_resync_args.get_redis_payload_func = RedisActPayload.redis_local_redo_dr.__name__
+            sub_pipelines.append(
+                {
+                    "act_name": _("{}-本地重建Slave").format(new_slave),
+                    "act_component_code": ExecuteDBActuatorScriptComponent.code,
+                    "kwargs": asdict(one_resync_args),
+                }
+            )
+        return sub_pipelines
+
     def redis_backend_scale_flow(self):
         """
         redis 扩缩容流程：
@@ -549,6 +589,18 @@ def redis_backend_scale_flow(self):
                         kwargs={**asdict(act_kwargs), **asdict(dns_kwargs)},
                     )
 
+            # > 4.0 版本需要重做一下slave，否则可能会丢数据
+            big_version = int(str.split(act_kwargs.cluster["db_version"], "-")[1])
+            if (
+                act_kwargs.cluster["cluster_type"]
+                in [
+                    ClusterType.TendisRedisInstance.value,
+                    ClusterType.TendisTwemproxyRedisInstance.value,
+                ]
+                and big_version >= 4
+            ):
+                sub_pipeline.add_parallel_acts(acts_list=self.redis_local_redo_dr(act_kwargs, sync_relations))
+
             sub_pipeline.add_act(act_name=_("Redis-人工确认"), act_component_code=PauseComponent.code, kwargs={})
 
             # 删除老实例的nodes域名