Additional Configuration for IBM BigInsights

If you are installing the SAS Embedded Process on IBM BigInsights 3.0, you must revise the hadoop.job.history.user.location property in the core-site.xml file that is in the SAS_HADOOP_CONFIG_PATH to a value other than the output directory. Otherwise, loading data into the Hive table fails. Here is an example where the output directory is set to /tmp.

<property>
    <name>hadoop.job.history.user.location</name>
    <value>/tmp</value>
</property>

If you are installing the SAS Embedded Process on IBM BigInsights 4.1, you must manually revise the following properties in the mapred-site.xml property file on the SAS client side. Otherwise, an error occurs when you submit a program to Hadoop.

The IBM BigInsights 4.1 mapred-site.xml property file contains numerous values with the parameter ${iop.version}, including mapreduce.application.classpath. You must change ${iop.version} to the actual cluster version. This example assumes that the current version is 4.1.0.0 and changes the mapreduce.admin.user.env property.

Change

<property>
    <name>mapreduce.admin.user.env</name>
    <value>/LD_LIBRARY_PATH=/usr/iop/${iop.version}/hadoop/lib/native</value>
</property>

to

<property>
    <name>mapreduce.admin.user.env</name>
    <value>/LD_LIBRARY_PATH=/usr/iop/4.1.0.0/hadoop/lib/native</value>
</property>