diff --git a/ChangeLog b/ChangeLog
index 8fb97d5..646cb1a 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -2,6 +2,7 @@
 	* VERSION: 1.6.17
 	* BUGFIX: transaction syncing panic when target redis is cluster. see
 	#145.
+	* IMPROVE: set max `sender.count` to 4096 if target redis type is cluster.
 2019-08-09 Alibaba Cloud.
 	* VERSION: 1.6.16
 	* BUGFIX: big key in `rump` mode all expired.
diff --git a/conf/redis-shake.conf b/conf/redis-shake.conf
index 281578f..1214700 100644
--- a/conf/redis-shake.conf
+++ b/conf/redis-shake.conf
@@ -187,8 +187,9 @@ heartbeat.network_interface =
 # 发送缓存的字节长度，超过这个阈值将会强行刷缓存发送
 sender.size = 104857600
 # sender flush buffer size of oplog number.
-# used in `sync`.
-# 发送缓存的报文个数，超过这个阈值将会强行刷缓存发送
+# used in `sync`. flush sender buffer when bigger than this threshold. for cluster,
+# the maximum value is 4096.
+# 发送缓存的报文个数，超过这个阈值将会强行刷缓存发送，如果目的端是cluster，这个最大只能是4096。
 sender.count = 4096
 # delay channel size. once one oplog is sent to target redis, the oplog id and timestamp will also
 # stored in this delay queue. this timestamp will be used to calculate the time delay when receiving