Hadoop 请求数据长度 Requested Data length 超过配置的最大值-CSDN博客

一、问题

现象

Spark 任务速度变慢也不失败。

DataNode 内存足够 CPU 负载不高 GC 时间也不长。

查看 DataNode 日志发现有些日志出现很多 Netty RPC 超时。超时的 destination 是一个 NameNode 节点然后查看 NameNode 节点的日志报错如下

二、解决方案

查找对应 Hadopo 源码

源码
org.apache.hadoop.ipc.Server.Connection#checkDataLength

    private void checkDataLength(int dataLength) throws IOException {
      if (dataLength < 0) {
        String error = "Unexpected data length " + dataLength +
                       "!! from " + getHostAddress();
        LOG.warn(error);
        throw new IOException(error);
      } else if (dataLength > maxDataLength) { 
        String error = "Requested data length " + dataLength +
              " is longer than maximum configured RPC length " + 
            maxDataLength + ".  RPC came from " + getHostAddress();
        LOG.warn(error);
        throw new IOException(error);    // <-------------- 异常从此处抛出来
      }
    }

this.maxDataLength = conf.getInt(CommonConfigurationKeys.IPC_MAXIMUM_DATA_LENGTH,
       CommonConfigurationKeys.IPC_MAXIMUM_DATA_LENGTH_DEFAULT);

  /** Max request size a server will accept. */
  public static final String IPC_MAXIMUM_DATA_LENGTH =
      "ipc.maximum.data.length";
  /** Default value for IPC_MAXIMUM_DATA_LENGTH. */
  public static final int IPC_MAXIMUM_DATA_LENGTH_DEFAULT = 64 * 1024 * 1024;

修改NameNode的hdfs-site.xml配置文件添加以下配置

<property>
  <name>ipc.maximum.data.length</name>
  <value>67108864</value>
  <description>This indicates the maximum IPC message length (bytes) that can be
    accepted by the server. Messages larger than this value are rejected by the
    immediately to avoid possible OOMs. This setting should rarely need to be
    changed.
  </description>
</property>

64M -> 256M

67108864 * 4 = 268435456

允许ipc通讯最大的数据包为256MB默认配置为64MB。

最后重启 NameNode再重启 DataNode。