[BIGTOP-1038] PIDFILE in service scripts of Hadoop is inconsistent with the path in hadoop-daemon.sh

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 0.6.0
• Component/s: Init scripts
• Assignee: Masatake Iwasaki
• Reporter: Masatake Iwasaki
• Created: 05/Aug/13 09:04
• Updated: 18/Sep/13 22:44

Description

PIDFILE in service scripts is fixed value because of variable expansion during packaging.

$ grep 'PIDFILE=' /etc/init.d/hadoop-*
/etc/init.d/hadoop-hdfs-datanode:PIDFILE="/var/run/hadoop-hdfs/hadoop-hdfs-datanode.pid"
/etc/init.d/hadoop-hdfs-namenode:PIDFILE="/var/run/hadoop-hdfs/hadoop-hdfs-namenode.pid"
/etc/init.d/hadoop-mapreduce-historyserver:PIDFILE="/var/run/hadoop-mapreduce/mapred-mapred-historyserver.pid"
/etc/init.d/hadoop-yarn-nodemanager:PIDFILE="/var/run/hadoop-yarn/yarn-yarn-nodemanager.pid"
/etc/init.d/hadoop-yarn-resourcemanager:PIDFILE="/var/run/hadoop-yarn/yarn-yarn-resourcemanager.pid"

However, HADOOP_IDENT_STRING composing pidfile path is configurable in hadoop-env.sh.

$ grep HADOOP_IDENT_STRING /usr/lib/hadoop/sbin/hadoop-daemon.sh
#   HADOOP_IDENT_STRING   A string representing this instance of hadoop. $USER by default
  export HADOOP_IDENT_STRING=$HADOOP_SECURE_DN_USER
if [ "$HADOOP_IDENT_STRING" = "" ]; then
  export HADOOP_IDENT_STRING="$USER"
  chown $HADOOP_IDENT_STRING $HADOOP_LOG_DIR
export HADOOP_LOGFILE=hadoop-$HADOOP_IDENT_STRING-$command-$HOSTNAME.log
log=$HADOOP_LOG_DIR/hadoop-$HADOOP_IDENT_STRING-$command-$HOSTNAME.out
pid=$HADOOP_PID_DIR/hadoop-$HADOOP_IDENT_STRING-$command.pid

As a result, the path of pidfile matches only when HADOOP_IDENT_STRING is service user name which is the default.

Issue Links

• is related to: BIGTOP-20 hard-coded pid location in Hadoop init.d scripts
• relates to: HADOOP-9979 HADOOP_IDENT_STRING should not be changed in hadoop-env.sh for hadoop daemons running as services