cpu/memory监控脚本的怪异现象

一开始nagios执行监控脚本的时候，总是不定期地偶尔地出现下面的错误

[: too many arguments

脚本一开始是这么写的：

guard_info=`ps aux | grep guard | grep -v grep | grep guard_replica_id | awk '{print $2}'`
echo guard_info:$guard_info
if [ "$guard_info" = "" ]; then
  echo "guard is not found"
  exit 0
fi
vsz=`echo $guard_info | awk '{print $3}'`
echo vsz:$vsz
cpu=`echo $guard_info | awk '{print $4}' | awk -F"." '{print $1}'`
echo cpu:$cpu


if [ $vsz -gt 606192 ]; then
  echo "vsz as large as $vsz"
  exit 2
elif [ $cpu -gt 5 ]; then
  echo "cpu percentage is $p_cpu"
  exit 2
else
  exit 0
fi

最后发现是guard本身会fork子进程，在调用execve之前和fork()之后有那么一个小时间窗口，这个时候会存在两个相同名字相同参数的进程存在。

解决办法就是用elapsed time来对grep出来的guard进程进行排序，取时间最长的那个。最后脚本变成这样：

guard_info=`ps -C guard -o pid=,etime=,vsz=,%cpu= --sort -etime | head -1`
echo guard_info:$guard_info
if [ "$guard_info" = "" ]; then
  echo "guard is not found"
  exit 0
fi
vsz=`echo $guard_info | awk '{print $3}'`
echo vsz:$vsz
cpu=`echo $guard_info | awk '{print $4}' | awk -F"." '{print $1}'`
echo cpu:$cpu


if [ $vsz -gt 606192 ]; then
  echo "vsz as large as $vsz"
  exit 2
elif [ $cpu -gt 5 ]; then
  echo "cpu percentage is $p_cpu"
  exit 2
else
  exit 0
fi