Hadoop HDFS文件經常使用操做及注意事項（更新）

1.Copy a file from the local file system to HDFS

The srcFile variable needs to contain the full name (path + file name) of the file in the local file system. 

The dstFile variable needs to contain the desired full name of the file in the Hadoop file system.

1 Configuration config = new Configuration();
2 FileSystem hdfs = FileSystem.get(config);
3 Path srcPath = new Path(srcFile);
4 Path dstPath = new Path(dstFile);
5 hdfs.copyFromLocalFile(srcPath, dstPath);

2.Create HDFS file

The fileName variable contains the file name and path in the Hadoop file system. 

The content of the file is the buff variable which is an array of bytes.

1 //byte[] buff - The content of the file 
2 //建立了一個HDFS文件，而且把buff數組的內容寫到了HDFS文件中。
3   Configuration config = new Configuration();
4   FileSystem hdfs = FileSystem.get(config);
5   Path path = new Path(fileName);
6   FSDataOutputStream outputStream = hdfs.create(path);
7   outputStream.write(buff, 0, buff.length);

3.Rename HDFS file

In order to rename a file in Hadoop file system, we need the full name (path + name) of 

the file we want to rename. The rename method returns true if the file was renamed, otherwise false.

1   Configuration config = new Configuration();
2   FileSystem hdfs = FileSystem.get(config);
3   Path fromPath = new Path(fromFileName);
4   Path toPath = new Path(toFileName);
5   boolean isRenamed = hdfs.rename(fromPath, toPath);

4.Delete HDFS file

In order to delete a file in Hadoop file system, we need the full name (path + name) 

of the file we want to delete. The delete method returns true if the file was deleted, otherwise false.

 1   Configuration config = new Configuration();
 2   FileSystem hdfs = FileSystem.get(config);
 3   Path path = new Path(fileName);
 4   boolean isDeleted = hdfs.delete(path, false);
 5 
 6 //Recursive delete:估計true是遞歸刪除該目錄下面的文件。
 7   Configuration config = new Configuration();
 8   FileSystem hdfs = FileSystem.get(config);
 9   Path path = new Path(fileName);
10   boolean isDeleted = hdfs.delete(path, true);

5.Get HDFS file last modification time

In order to get the last modification time of a file in Hadoop file system, 

we need the full name (path + name) of the file.

1   Configuration config = new Configuration();
2   FileSystem hdfs = FileSystem.get(config);
3   Path path = new Path(fileName);
4   FileStatus fileStatus = hdfs.getFileStatus(path);
5   long modificationTime = fileStatus.getModificationTime

6.Check if a file exists in HDFS

In order to check the existance of a file in Hadoop file system, 

we need the full name (path + name) of the file we want to check. 

The exists methods returns true if the file exists, otherwise false.

1 Configuration config = new Configuration();
2 FileSystem hdfs = FileSystem.get(config);
3 Path path = new Path(fileName);
4 boolean isExists = hdfs.exists(path);

7.Get the locations of a file in the HDFS cluster

 A file can exist on more than one node in the Hadoop file system cluster for two reasons:

Based on the HDFS cluster configuration, Hadoop saves parts of files on different nodes in the cluster.

Based on the HDFS cluster configuration, Hadoop saves more than one copy of each file on different nodes for redundancy (The default is three).

 1 Configuration config = new Configuration();
 2 FileSystem hdfs = FileSystem.get(config);
 3 Path path = new Path(fileName);
 4 FileStatus fileStatus = hdfs.getFileStatus(path);
 5 BlockLocation[] blkLocations = hdfs.getFileBlockLocations(fileStatus, 0, fileStatus.getLen());
 6 int blkCount = blkLocations.length;
 7 for (int i = 0; i < blkCount; i++) {
 8     String[] hosts = blkLocations[i].getHosts();
 9     // Do something with the block hosts
10 }

8. Get a list of all the nodes host names in the HDFS cluster

  his method casts the FileSystem Object to a DistributedFileSystem Object. 

  This method will work only when Hadoop is configured as a cluster. 

  Running Hadoop on the local machine only, in a non cluster configuration will  cause this method to throw an Exception.

1   Configuration config = new Configuration();
2   FileSystem fs = FileSystem.get(config);
3   DistributedFileSystem hdfs = (DistributedFileSystem) fs;
4   DatanodeInfo[] dataNodeStats = hdfs.getDataNodeStats();
5   String[] names = new String[dataNodeStats.length];
6   for (int i = 0; i < dataNodeStats.length; i++) {
7       names[i] = dataNodeStats[i].getHostName();
8   }

 

遇到的問題及解決：

1.文件append的問題

hadoop的版本1.0.4之後，API中已經有了追加寫入的功能，但不建議在生產環境中使用，緣由以下：Does HDFS allow appends to files? This is currently set to false because there are bugs in the "append code" and is not supported in any prodction cluster.
若是要測試的話，須要把dfs.support.appen的參數設置爲true，否則客戶端寫入的時候會報錯：

Exception in thread "main" org.apache.hadoop.ipc.RemoteException: java.io.IOException: Append to hdfs not supported. Please refer to dfs.support.append configuration parameter.

解決：修改namenode節點上的hdfs-site.xml。

  <property>
    <name>dfs.support.append</name>
    <value>true</value>
  </property>

2.在Hadoop-1.0.4和Hadoop-2.2的使用append時，需求：追加寫入文件，若是文件不存在，需求先建立。

異常：

Exception in thread "main" org.apache.hadoop.ipc.RemoteException: org.apache.hadoop.hdfs.protocol.AlreadyBeingCreatedException: failed to create file /huangq/dailyRolling/mommy-dailyRolling for DFSClient_-1456545217 on client 10.1.85.243 because current leaseholder is trying to recreate file.
    at org.apache.hadoop.hdfs.server.namenode.FSNamesystem.recoverLeaseInternal(FSNamesystem.java:1374)
    at org.apache.hadoop.hdfs.server.namenode.FSNamesystem.startFileInternal(FSNamesystem.java:1246)
    at org.apache.hadoop.hdfs.server.namenode.FSNamesystem.appendFile(FSNamesystem.java:1426)
    at org.apache.hadoop.hdfs.server.namenode.NameNode.append(NameNode.java:643)
    at sun.reflect.GeneratedMethodAccessor25.invoke(Unknown Source)
    at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:25)
    at java.lang.reflect.Method.invoke(Method.java:597)

代碼版本1：（出現上述報錯的代碼）

1 FileSystem fs = FileSystem.get(conf);
2 Path dstPath = new Path(dst);
3 if (!fs.exists(dstPath)) {
4     fs.create(dstPath);
5 } 
6 FSDataOutputStream fsout = fs.append(dstPath);
7 BufferedWriter bw = new BufferedWriter(new OutputStreamWriter(fsout));

異常的緣由：FSDataOutputStream create(Path f) 產生了一個輸出流，建立完後須要關閉。

解決：建立完文件以後，關閉流FSDataOutputStream。

FileSystem fs = FileSystem.get(conf);
Path dstPath = new Path(dst);
if (!fs.exists(dstPath)) {
    fs.create(dstPath).close();
} 
FSDataOutputStream fsout = fs.append(dstPath);