HDFS Application Development Rules

Set the HDFS NameNode metadata storage path

NameNode metadata is stored in ${BIGDATA_DATA_HOME}/namenode/data by default. This parameter sets the storage path of HDFS metadata.

Enable NameNode image backup for the HDFS

fs.namenode.image.backup.enable specifies whether to enable the NameNode image backup function. You need to set this parameter to true. Then the system can periodically back up the NameNode data.

Set the HDFS DataNode data storage path

DataNode data is stored in ${BIGDATA_DATA_HOME}/hadoop/dataN/dn/datadir by default. N indicates the number of directories is greater than or equal to1.

For example, ${BIGDATA_DATA_HOME}/hadoop/data1/dn/datadir, ${BIGDATA_DATA_HOME}/hadoop/data2/dn/datadir.

After the storage path is set, data is stored in the corresponding directory of each mounted disk on a node.

Improve HDFS read/write performance

The data write process is as follows:

After receiving service data and obtaining the data block number and location from the NameNode, the HDFS client contacts DataNodes and establishes a pipeline with the DataNodes to be written. Then, the HDFS client writes data to DataNode1 using a proprietary protocol, and DataNode1 writes data to DataNode2 and DataNode3 (three duplicates). After data is written, a message is returned to the HDFS client.

1. Set a proper block size. For example, set dfs.blocksize to 268435456 (256 MB).
2. It is not necessary to cache the big data that is not reused. In this case, set the following parameters to false:

   dfs.datanode.drop.cache.behind.reads and dfs.datanode.drop.cache.behind.writes

Set the MapReduce intermediate file storage path

Only one default path is provided for storing MapReduce intermediate files, that is, ${hadoop.tmp.dir}/mapred/local.It is recommended that intermediate files be stored on each disk.

For example, /hadoop/hdfs/data1/mapred/local, /hadoop/hdfs/data2/mapred/local, /hadoop/hdfs/data3/mapred/local. Directories that do not exist are automatically ignored.

Release applied resources in finally during Java development.

Applied HDFS resources are released in try/finally and cannot be released outside the try statement only. Otherwise, resource leakage occurs.

HDFS file operation APIs

Almost all Hadoop file operation classes are in the org.apache.hadoop.fs package. These APIs support operations such as opening, reading, writing, and deleting a file. FileSystem is the interface class provided for users in the Hadoop class library. FileSystem is an abstract class. Concrete classes can be obtained only using the get method. The get method has multiple overload versions, and the following get method is often used.

static FileSystem get(Configuration conf);

This class encapsulates almost all file operations, such as mkdir and delete. The program library framework for file operations is as follows:

operator() 
{
     Obtain the Configuration object.
     Obtain the FileSystem object.
     Perform file operations.
 } 

HDFS initialization method

HDFS initialization is a prerequisite for using APIs provided by HDFS.

To initialize HDFS, load the HDFS service configuration file, implement Kerberos security authentication, and instantiate FileSystem. Obtain keytab files for Kerberos security authentication in advance.

Example:
private  void init() throws IOException {
 Configuration conf = new Configuration();
 // Read a configuration file.
  conf.addResource("user-hdfs.xml");
 // Implement security authentication in security mode.
  if ("kerberos".equalsIgnoreCase(conf.get("hadoop.security.authentication"))) {
 String PRINCIPAL = "username.client.kerberos.principal";
 String KEYTAB = "username.client.keytab.file";
 // Set the keytab key file.
  conf.set(KEYTAB, System.getProperty("user.dir") + File.separator + "conf" + File.separator + conf.get(KEYTAB));
  // Set the Kerberos configuration file path. */
 String krbfilepath = System.getProperty("user.dir") + File.separator + "conf" + File.separator + "krb5.conf";
 System.setProperty("java.security.krb5.conf", krbfilepath);
 // Implement login authentication. */
 SecurityUtil.login(conf, KEYTAB, PRINCIPAL);
 }
  // Instantiate FileSystem.
 fSystem = FileSystem.get(conf);
 } 

Upload local files to the HDFS

FileSystem.copyFromLocalFile (Path src, Patch dst) is used to upload local files to a specified directory in the HDFS. src and dst indicate complete file paths.

Example:

public class CopyFile {
        public static void main(String[] args) throws Exception { 
        Configuration conf=new Configuration();
        FileSystem hdfs=FileSystem.get(conf);
              //Local file
                 Path src =new Path("D:\\HebutWinOS");
                  //To the HDFS
        Path dst =new Path("/"); 
               hdfs.copyFromLocalFile(src, dst);
        System.out.println("Upload to"+conf.get("fs.default.name"));
              FileStatus files[]=hdfs.listStatus(dst);
              for(FileStatus file:files){ 
                 System.out.println(file.getPath());
              }
        }
 } 

Create files on the HDFS

FileSystem.mkdirs (Path f) is used to create folders on HDFS. f indicates a complete folder path.

Example:

public class CreateDir { 
    public static void main(String[] args) throws Exception{
         Configuration conf=new Configuration();
         FileSystem hdfs=FileSystem.get(conf);
                Path dfs=new Path("/TestDir");
                hdfs.mkdirs(dfs);
      }
  } 

Query the modification time of an HDFS file

FileSystem.getModificationTime() is used to query the modification time of a specified HDFS file.

Example:

     public static void main(String[] args) throws Exception {
          Configuration conf=new Configuration();
          FileSystem hdfs=FileSystem.get(conf);
          Path fpath =new Path("/user/hadoop/test/file1.txt");
          FileStatus fileStatus=hdfs.getFileStatus(fpath);
          long modiTime=fileStatus.getModificationTime();
          System.out.println("file1.txt modification time is"+modiTime); 
    }

Read all files in an HDFS directory

FileStatus.getPath() is used to query all files in an HDFS directory.

Example:

public static void main(String[] args) throws Exception {
         Configuration conf=new Configuration();
         FileSystem hdfs=FileSystem.get(conf);
         Path listf =new Path("/user/hadoop/test");

         FileStatus stats[]=hdfs.listStatus(listf);
         for(int i = 0; i < stats.length; ++i) {
            System.out.println(stats[i].getPath().toString());
         }         
         hdfs.close();     
  } 

Query the location of a specified file in an HDFS cluster

FileSystem.getFileBlockLocation (FileStatus file, long start, long len) is used to query the location of a specified file in an HDFS cluster. file indicates a complete file path, and start and len specify the file path.

Example:

public static void main(String[] args) throws Exception {
         Configuration conf=new Configuration();
         FileSystem hdfs=FileSystem.get(conf);
         Path fpath=new Path("/user/hadoop/cygwin");

         FileStatus filestatus = hdfs.getFileStatus(fpath);
         BlockLocation[] blkLocations = hdfs.getFileBlockLocations(filestatus, 0, filestatus.getLen());

         int blockLen = blkLocations.length;
         for(int i=0;i < blockLen;i++){
             String[] hosts = blkLocations[i].getHosts();
             System.out.println("block_"+i+"_location:"+hosts[0]); 
         }
    } 

Obtain all node names in an HDFS cluster

DatanodeInfo.getHostName() is used to obtain all node names in an HDFS cluster.

Example:

public static void main(String[] args) throws Exception {         
        Configuration conf=new Configuration();         
        FileSystem fs=FileSystem.get(conf);
        
        DistributedFileSystem hdfs = (DistributedFileSystem)fs;

        DatanodeInfo[] dataNodeStats = hdfs.getDataNodeStats(); 
                
        for(int i=0;i < dataNodeStats.length;i++){            
            System.out.println("DataNode_"+i+"_Name:"+dataNodeStats[i].getHostName());         
        }     
   } 

Multithread security login mode

If multiple threads are performing login operations, the relogin mode must be used for the subsequent logins of all threads after the first successful login of an application.

login example code:

  private Boolean login(Configuration conf){     
    boolean flag = false;     
    UserGroupInformation.setConfiguration(conf);          
    try {       
          UserGroupInformation.loginUserFromKeytab(conf.get(PRINCIPAL), conf.get(KEYTAB));       
          System.out.println("UserGroupInformation.isLoginKeytabBased(): " +UserGroupInformation.isLoginKeytabBased());       
          flag = true;     
        } catch (IOException e) {       
           e.printStackTrace();     
        }     
     return flag;        
  }

relogin example code:

public Boolean relogin(){         
        boolean flag = false;         
        try {                        
              UserGroupInformation.getLoginUser().reloginFromKeytab();    
              System.out.println("UserGroupInformation.isLoginKeytabBased(): " +UserGroupInformation.isLoginKeytabBased());           
              flag = true;         
            } catch (IOException e) {             
                e.printStackTrace();         
            }         
        return flag;     
   }

Repetitive logins will cause a newly created session to overwrite the previous session. As a result, the previous session cannot be maintained or monitored, and some functions are unavailable after the previous session expires.