The init.d scripts have a hard-coded PID file, e.g.:
The init scripts should instead respect the values set in hadoop-env.sh
Current work-around is to set HADOOP_PID_DIR in hadoop-env.sh to /var/run/hadoop-0.20. Note that HADOOP_IDENT_STRING must be set to the same value in both /etc/default/hadoop-0.20 and hadoop-env.sh for the pid file to be found. It defaults to "hadoop" in /etc/default/hadoop-0.20.