Common HDFS Commands

Start or Stop Hadoop

To start or stop SAS High-Performance Deployment for Hadoop, log on to the machine that is used as the NameNode. Log on with the user ID that was selected as the service account (that account is often named hdfs).
  • The start command is as follows:
    /hadoop-installation-directory/sbin/start-dfs.sh
  • The stop command is as follows:
    /hadoop-installation-directory/sbin/stop-dfs.sh
Note: A typical installation directory for the SAS High-Performance Deployment for Hadoop is /hadoop/hadoop-version.

Create and Protect Directories

To create and manage access to directories in HDFS, log on to the machine that hosts the NameNode, and use the hadoop command.
Note: To get started, use the hdfs user account to create and manage directories. Once the beginning of a directory structure is created and permissions are changed, other user accounts can be used to manage access to directories.
To create a general purpose directory:
  1. As the hdfs user account, create a directory named /shared:
    ./hadoop fs -mkdir /shared
  2. Open up access permissions on the directory:
    ./hadoop fs -chmod 1777 /shared
    Note: This permissions mode enables only the superuser, directory owner, and file owner to delete or move files within the directory.
  3. Confirm that the commands succeeded:
    ./hadoop fs -ls /
    
    Found 3 items
    drwxr-xr-x   - hdfs supergroup            0 2014-02-03 21:38 /data
    drwxrwxrwt   - hadoop supergroup          0 2014-02-14 21:23 /shared
    drwxrwxrwt   - hdfs supergroup            0 2014-01-17 11:07 /tmp
    drwxr-xr-x   - hdfs supergroup            0 2014-02-13 08:45 /user
    
To set up a directory for members of the sales group:
  1. Create a directory named /dept/sales:
    ./hadoop fs -mkdir -p /dept/sales
  2. Change the group ID:
    ./hadoop fs -chgrp sales /dept/sales
    Note: The preceding command assumes that an operating system group that is named sales exists. You can use the SAS High-Performance Computing Management Console to create the group on the machines in the cluster. After you create the group, stop and then start Hadoop (so that the group is recognized).
  3. Provide access to only the hdfs user account and members of the sales group:
    ./hadoop fs -chmod 770 /dept/sales
  4. Confirm that the commands succeeded:
    ./hadoop fs -ls /dept
    
    Found 1 items
    drwxrwx---   - hdfs sales            0 2014-02-14 21:29 /dept/sales
    
Note: The HDFS directory structure is similar to a UNIX file system. Directories have a user ID, group ID, and associated access permissions. More information about the hadoop command is available from http://hadoop.apache.org.