Client Configuration

These instructions are for Hadoop distros other than MapR, if you are using MapR go to the Configure Pentaho for MapR page.

Kettle Client

1. Download and extract Kettle CE from the Downloads page.
   The Kettle Client comes pre-configured for Apache Hadoop 0.20.2. If you are using this distro and version, no further configuration is required.
2. Configure PDI Client for a different version of Hadoop
   1. Delete $PDI_HOME/libext/pentaho/hadoop-0.20.2-core.jar
   2. For all other distributions you should replace this core jar with the one from your cluster. For example, if you are using Cloudera CDHu2:
      Copy $HADOOP_HOME/hadoop-core-0.20.2-cdh3u2.jar to $PDI_HOME/libext/pentaho
   3. For certain Hadoop distributions or versions (for example Hadoop 0.20.205) you also need to have Apache Commons Configuration included in your set of PDI libraries.  In that case copy commons-configuration-1.7.jar to $PDI_HOME/libext/commons
   4. For certain Hadoop distributions or versions (for example Cloudera CDH3 Update 3) you also need to copy $HADOOP_HOME/lib/guava-r09-jarjar.jar to $PDI_HOME/libext/pentaho.

Pentaho Report Designer (PRD)

1. Download and extract PRD from the Downloads page.
   The PRD comes pre-configured for Apache Hadoop 0.20.2. If you are using this distro and version, no further configuration is required.
2. Configure PRD for a different version of Hadoop
   1. Delete $PRD_HOME/lib/jdbc/hadoop-0.20.2-core.jar
   2. Copy $HADOOP_HOME/hadoop-core.jar from your distribution into $PRD_HOME/lib/jdbc
   3. Copy commons-configuration-1.7.jar to $PRD_HOME/lib (in case this is needed for certain Hadoop variants)
   4. For certain Hadoop distributions or versions (for example Cloudera CDH3 Update 3) you also need to copy $HADOOP_HOME/lib/guava-r09-jarjar.jar to $PRD_HOME/lib.

Hadoop Node Configuration

All TaskTracker nodes must have the pentaho-mapreduce (PHD) package installed on them. Perform the following steps:

1. Download the latest PHD archive from the Downloads page.
2. Extract the PHD archive into /opt/pentaho/pentaho-mapreduce
3. Update $HADOOP_HOME/conf/hadoop-env.sh and add /opt/pentaho/pentaho-mapreduce/lib/* to the HADOOP_CLASSPATH

   export HADOOP_CLASSPATH=/opt/pentaho/pentaho-mapreduce/lib/*:$HADOOP_CLASSPATH
   

4. Update $HADOOP_HOME/conf/mapred-site.xml and add the following properties if they do not exist:

   1. <property>
        <name>pentaho.kettle.home</name>
        <value>/opt/pentaho/pentaho-mapreduce</value>
      </property>
      
      <property>
        <name>pentaho.kettle.plugins.dir</name>
        <value>/opt/pentaho/pentaho-mapreduce/plugins</value>
      </property>
      

   2. Note: pentaho.kettle.home should contain a .kettle/ folder with kettle.properties. A shared.xml can also be placed here if you require any Kettle shared objects in the transformations execution within Pentaho MapReduce. These files will be created automatically if the user the TaskTracker process is executed as has write access to pentaho.kettle.home. If not, you should manually create at least .kettle/kettle.properties.
5. Restart the Hadoop JobTracker and TaskTracker nodes (usually stop-mapred/start-mapred will do the trick) for the HADOOP_CLASSPATH change to take effect.