一个以前用的提交hadoop作业的shell脚本

做个笔记，用到的时候可以看一看。感谢大师兄！怀念校园那段日子。

#!/bin/sh
arg0=$0
arg1=$1
arg2=$2
arg3=$3
arg4=$4
arg5=$5
JAR=MyJob3-1.jar
JOBNAME=MyJob3
echo $#
args=$#

usage()
{
    echo "Usage: $arg0 <nJobs> <s interval> <MapTask> <outputfile> [dataOverlap file number]"
}
run()
{
    if [ $args -lt 4 ]; then
        usage
        exit
    fi
    HADOOP_STATUS=`hadoop job -list|grep -U1 JobId`
    JOBS=`echo $HADOOP_STATUS|awk '{print $2}'`
    while [ $JOBS != "jobs:0" ]; do
       sleep 10
    HADOOP_STATUS=`hadoop job -list|grep -U1 JobId`
    JOBS=`echo $HADOOP_STATUS|awk '{print $2}'`
    done
    
    echo "TEST: Launch $arg1 jobs ($JOBNAME) using jar $JAR with $arg2 seconds interval. Input file is $arg3 and Output file is $arg4"  
    hadoop fs -rmr output*
    if [ ! -d  $arg4 ]; then
        mkdir -p $arg4
    fi
    if [ $args -eq 5 ]; then
        file=$arg5
    else
        file=6
    fi
    COUNT=0
    for i in `seq $arg1`; do
        index=$[$i%$file]
    if [ $index -eq 0 ]; then
        index=$file
    fi    
    echo "hadoop jar $JAR $JOBNAME access$index-$arg3.log output$i"
    (time `hadoop jar $JAR $JOBNAME access$index-$arg3.log output$i >/dev/null  2>&1`)  2>$arg4/log$i &
#    (time `hadoop jar $JAR $JOBNAME access1-$arg3.log output$i >/dev/null  2>&1`)  2>$arg4/log$i &
      COUNT=`expr $COUNT + 1`
    sleep $arg2
    done
}

run