When you use the Hadoop server in SAS^® Visual Analytics with Distributed SAS^® LASR™ system, you might see some errors when you update the Hadoop server or view the SAS Visual Analytics report, or load or unload tables on SAS LASR Analytic Server.

The error messages that you see in SAS Visual Analytics might look like this:

ERROR: The datanode xxxxxx could not write a file block to HDFS. This might be due to filling up the partition where HDFS resides.

ERROR: Communication failure between LASR Analytic Server and client application.

ERROR: An I/O error has occurred on file xxxx.xxxx.DATA.

The error in the Hadoop namenode log file might look like this:

FATAL org.apache.hadoop.hdfs.server.namenode.NameNode: Exception in namenode join 
java.lang.OutOfMemoryError: Java heap space

The root cause of the problem is that the Java Heap size is too low on Hadoop. The default value for JAVA_HEAP_MAX is 1024M. SAS recommends raising the size to 4096M.

Workaround

1. Modify the script file hadoop-config.sh under /hadoop-2.4.0/bin/hdfs.
2. Find the if command = namenode and same for secondary name node in this script and change the current value to Xmx4096m for JAVA_HEAP_MAX.

Here is an example:

if [ "$COMMAND" = "namenode" ] ; then ..... JAVA_HEAP_MAX=-Xmx4096m HADOOP_OPTS="$HADOOP_OPTS $HADOOP_NAMENODE_OPTS" ......... elif [ "$COMMAND" = "secondarynamenode" ] ; then ... JAVA_HEAP_MAX=-Xmx4096m ...
3. Stop and restart Hadoop (at least just the name node) with the stop-dfs.sh and start-dfs.sh script in order to pick up the new settings.

Operating System and Release Information