HDFS DATANODE 磁盤容量的最小值

HDFS的DATANODE的剩餘空間具體要到多大？關於這個問題，下面記錄下對這個問題的調查 

昨天，討論羣裏面給出了一個異常： 

op@odbtest bin]$ hadoop fs -put ../tmp/file3 /user/hadoop/in2
14/01/15 02:14:09 WARN hdfs.DFSClient: DataStreamer Exception
org.apache.hadoop.ipc.RemoteException(java.io.IOException): File /user/hadoop/in2/file3._COPYING_ could only be replicated to 0 nodes instead of minReplication (=1).  There are 1 datanode(s) running and no node(s) are excluded in this operation.
        at org.apache.hadoop.hdfs.server.blockmanagement.BlockManager.chooseTarget(BlockManager.java:1384)
        at org.apache.hadoop.hdfs.server.namenode.FSNamesystem.getAdditionalBlock(FSNamesystem.java:2477)
        at org.apache.hadoop.hdfs.server.namenode.NameNodeRpcServer.addBlock(NameNodeRpcServer.java:555)
        at org.apache.hadoop.hdfs.protocolPB.ClientNamenodeProtocolServerSideTranslatorPB.addBlock(ClientNamenodeProtocolServerSideTranslatorPB.java:387)
        at org.apache.hadoop.hdfs.protocol.proto.ClientNamenodeProtocolProtos$ClientNamenodeProtocol$2.callBlockingMethod(ClientNamenodeProtocolProtos.java:59582)
        at org.apache.hadoop.ipc.ProtobufRpcEngine$Server$ProtoBufRpcInvoker.call(ProtobufRpcEngine.java:585)
        at org.apache.hadoop.ipc.RPC$Server.call(RPC.java:928)
        at org.apache.hadoop.ipc.Server$Handler$1.run(Server.java:2048)
        at org.apache.hadoop.ipc.Server$Handler$1.run(Server.java:2044)
        at java.security.AccessController.doPrivileged(Native Method)
        at javax.security.auth.Subject.doAs(Subject.java:396)
        at org.apache.hadoop.security.UserGroupInformation.doAs(UserGroupInformation.java:1491)
        at org.apache.hadoop.ipc.Server$Handler.run(Server.java:2042)

該異常的只在NN的日誌中拋出，而DN中沒有相關內容，這說明，這是在NN進行block分配的時候作了檢查。 

這種狀況通常都是DATANODE 變成dead，或者是datanode的磁盤容量不夠了。 

因此建議問題提出者，給DN的datadir空出一部分空間以後，操做正常 


可是，該問題的提出者，給出report 數據： 

[hadoop@odbtest bin]$ hdfs dfsadmin -report
Configured Capacity: 8210259968 (7.65 GB)
Present Capacity: 599728128 (571.95 MB)
DFS Remaining: 599703552 (571.92 MB)
DFS Used: 24576 (24 KB)
DFS Used%: 0.00%
Under replicated blocks: 0
Blocks with corrupt replicas: 0
Missing blocks: 0

-------------------------------------------------
Datanodes available: 1 (1 total, 0 dead)

Live datanodes:
Name: 192.168.136.128:50010 (odbtest)
Hostname: odbtest
Decommission Status : Normal
Configured Capacity: 8210259968 (7.65 GB)
DFS Used: 24576 (24 KB)
Non DFS Used: 7610531840 (7.09 GB)
DFS Remaining: 599703552 (571.92 MB)
DFS Used%: 0.00%
DFS Remaining%: 7.30%
Last contact: Tue Jan 14 23:47:26 PST 2014

按照report的數據DFS還剩下(571.92 MB)的大小，應該是能夠建立的，可是拋出了這個異常，確定是對DATANODE的剩餘最小容量作了限制。查了一下HADOOP 2.2.0的源碼， 
org.apache.hadoop.hdfs.server.blockmanagement.BlockPlacementPolicyDefault的方法isGoodTarget中，會對DATANODE的剩餘容量進行判斷： 

long remaining = node.getRemaining() -
                   (node.getBlocksScheduled() * blockSize);
  // check the remaining capacity of the target machine
  if (blockSize* HdfsConstants.MIN_BLOCKS_FOR_WRITE>remaining) {
    if(LOG.isDebugEnabled()) {
      threadLocalBuilder.get().append(node.toString()).append(": ")
        .append("Node ").append(NodeBase.getPath(node))
        .append(" is not chosen because the node does not have enough space ");
    }
    return false;
  }

代碼中說了，當剩餘容量小於blockSize* HdfsConstants.MIN_BLOCKS_FOR_WRITE的時候，會返回false，而默認狀況下 blockSize* HdfsConstants.MIN_BLOCKS_FOR_WRITE=128M*5=640M> 571.92 MB,這就解釋了這個異常發生的緣由。