VASMP Procedure

Example: Copying Tables from One Hadoop Installation to Another


This example does not apply to a non-distributed SAS LASR Analytic Server. It might be necessary to work with more than one Hadoop installation so that you can copy SASHDAT files from one Hadoop installation to a newer version. The SAS LASR Analytic Server must be co-located with both Hadoop installations and both versions of Hadoop must be running.
Note: Using the HADOOPHOME= option to switch between Hadoop installations is a server-wide change. If users access the server while the setting is being switched, they might accidentally access the older Hadoop installation. Consider starting a server for the exclusive use of copying files.


proc lasr create port=12636 serverpermissions=700; 1
    performance host="" install="/opt/TKGrid" nodes=all;

libname private sasiola host="" port=12636 tag='hps';

data private.iris; set sashelp.iris; run;  /* a table must be active */

proc VASMP data=private.iris; 2
    serverparm hadoophome="/olderhadoop/path"; 3

proc lasr add hdfs(path="/dept/sales/y2011" direct) port=12636; 4
    performance host="";

proc VASMP data=private.y2011(tag="dept.sales"); 5
    serverparm hadoophome="/newerhadoop/path"; 6
    save path="/dept/sales/"; 7

Program Description

  1. Starting a server with SERVERPERMISSIONS=700 creates a single-user server. This is not required but can be used to prevent users from accessing the server while the HADOOP_HOME value is changed and accidentally accessing older or incorrect data.
  2. You must have an active table. You can specify an active table with the DATA= option. Any table, such as the Iris data set can be used.
  3. Use the SERVERPARM statement to specify the path to the older Hadoop installation with the HADOOPHOME= option. Specify the same path that is returned for the HADOOP_HOME environment variable for the older installation. Example: /hadoop/hadoop-0.21.
  4. You must specify the DIRECT option. This statement loads table y2011 into memory from the /dept/sales directory in HDFS.
  5. The TAG= option must be used to specify the in-memory table. The server tag matches the HDFS path to the table, but the slashes are replaced with periods (.). If the table was loaded from /, then specify TAG=HADOOP.
  6. Use the SERVERPARM statement to specify the path to the newer Hadoop installation. Example: /hadoop-0.23/hadoop-0.23.1.
  7. The SAVE statement writes the y2011 table to HDFS in the /dept/sales directory. The HDFS directory is in the newer Hadoop installation.