About datanode direct access problems ?

hdfs file is the mechanism to take the client through the namenode get metadata information , and then extract the data directly connected datanode ;
It is not to be directly connected datanode datanode data on the whole do down ? This is no problem with it ?
initial contact hadoop hope prawn cordial guidance
------ Solution ---------------------------- ----------------
hdfs the file is divided into one or more blocks to store , these blocks are not necessarily the same data are stored in the node , namenode recorded each block in each file where the data node information , so you do not go through namenode datanode taken down even from the data , do not know how to reconstruct the data into files.
------ Solution ---------------------------------------- ----
DN is RPC interface , you can visit !
------ For reference only -------------------------------------- -
Thank you for reminding , oh pro
------ For reference only ---------------------------- -----------


positive solution