diff --git a/docs/en/hadoop_java_sdk.md b/docs/en/hadoop_java_sdk.md
index 43ed0c1141e422d21daa05b59815d85396f791b0..19d48fbe411b68b07b2887e709b6f6ce07a69d1f 100644
--- a/docs/en/hadoop_java_sdk.md
+++ b/docs/en/hadoop_java_sdk.md
@@ -30,6 +30,8 @@ JuiceFS provides [Hadoop-compatible FileSystem](https://hadoop.apache.org/docs/c
 > **NOTICE**:
 >
 > JuiceFS use local mapping of user and UID. So, you should [sync all the needed users and their UIDs](sync_accounts_between_multiple_hosts.md) across the whole Hadoop cluster to avoid permission error. You can also specify a global user list and user group file, please refer to the [relevant configurations](#other-configurations).
+>
+> JuiceFS Hadoop Java SDK need extra 4 * ``juicefs.memory-size`` off-heap memory at most. By default, up to 1.2 GB of additional memory is required (depends on write load).
 
 ## Hadoop Compatibility
 
diff --git a/docs/zh_cn/hadoop_java_sdk.md b/docs/zh_cn/hadoop_java_sdk.md
index e9596215e65cd156f699323d475513584406315c..3d34d166131a861eb074450411c7666b6a5f8d88 100644
--- a/docs/zh_cn/hadoop_java_sdk.md
+++ b/docs/zh_cn/hadoop_java_sdk.md
@@ -30,6 +30,8 @@ JuiceFS 提供兼容 HDFS 接口的 Java 客户端来支持 Hadoop 生态中的
 > **注意**：
 >
 > 由于 JuiceFS 默认使用本地的 user 和 UID 映射。因此，在分布式环境下使用，需要[同步所有需要使用的 user 和 UID](sync_accounts_between_multiple_hosts.md) 到所有的 Hadoop 节点上，以避免权限问题。也可以指定一个全局的用户列表和所属用户组文件，具体请参见[相关配置](#其他配置)。
+> 
+> JuiceFS Hadoop Java SDK 最多需要额外使用 4 * ``juicefs.memory-size`` 的 off-heap 内存用来加速读写性能，默认情况下，最多需要额外 1.2 GB 内存（取决于写入负载）。
 
 ## Hadoop 兼容性