Including HIVE-6259, HIVE-6310, etc., recent test failures of minimr tests
are seemed to be related to memory issue (not all of them but mostly).
{noformat}
java.lang.Throwable: Child Error
at org.apache.hadoop.mapred.TaskRunner.run(TaskRunner.java:271)
Caused by: java.io.IOException: Task process exit with nonzero status of 1.
at org.apache.hadoop.mapred.TaskRunner.run(TaskRunner.java:258)
{noformat}
Currently we uses 4 datanode + 4 task tracker mini cluster. If we cannot
reserve more memory only for hive tests (it would be really appreciated if
it could be), should we use smaller mini cluster for pre-tests?
