Hoot, these are big numbers - some thoughts 1) does your machine have 1000GB to spare for each java child thread (each mapper + each reducer)? mapred.child.java.opts / -Xmx1048576m 2) does each of your daemons need / have 10G? HADOOP_HEAPSIZE=10000
hth EF > # The maximum amount of heap to use, in MB. Default is 1000. > export HADOOP_HEAPSIZE=10000 > > <name>mapred.child.java.opts</name> > <value>-Xmx1048576m</value> > </property> > >
