Hi there folks,

I'm newbie here and need some help from the gurus in this forum, pls excuse me as my english is not very good.


WARNING long post!!

I have a simple script running on my box which basically polls every 5mins to check the utilization of certain iplanet SOP application processes, if the process exceed > 80% it gets killed and the application will automatically respawn a new process but running at lower cpu.

while true
do
rm -f /tmp/HIGH_SOP_CPU.*
ps -edf -o user,pcpu,pid | grep -i sopias | awk '$2 > 80.0 { print }' >> /tmp/HIGH_SOP_CPU.`date +%d%m%y%H%M`
if (test -s /tmp/HIGH_SOP_CPU.`date +%d%m%y%H%M`) then
kill -9 `ps -edf -o user,pcpu,pid | grep -i sopias | awk '$2 > 80.0 { print $3 }
'`
fi
sleep 300
done


However the problem is this script only checks for a single process > 80%, if there are more than 1 or a few sopias process running concurrently and all of them under 80% utilization but collectively contribute to taking > 80% of system cpu resources it goes undetected by the script until the users complain then it needs to be killed manually.

I'm trying to either expand the current script or write a seperate one which greps all the sopias processes, find the total utilization and if > 80%, find the largest utilized process and kill it. Of course it's easier said than done and I'm still trying to find out how to do it.

I'm trying to find out how I can couple these 2 awk statements together without using flatfiles.

awk '{for (i=1; i<=NF; i++) s=s+$i}; END {print s}'` `ps -edf -o user,pcpu,pid | grep -i sopias | awk '{ print $2}'`

basically the ps statement greps a list of all the utilization and pipes to the for-loop which does the calculations, if I use a flatfile for the `ps` it works fine.

any ideas how not to do it with a flatfile? And maybe a sample awk script that can show me how to find the largest number from a range of numbers.

thx in advance

use the following logic in your scripts

df -k |grep %|grep -v "/proc"|grep -v "/stats"|grep -v "/dev/fd"|sed -e 's/%//'|awk '{ print $6 " " $5 }' > /tmp/tmpfile
cat /tmp/tmpfile | while read dsk usg
do
        if [ $usg -gt 99 ]
        then
                echo "$dsk partition is $usg% full" >> $FILETMP2
        fi
done

hi thx for your reply.

my original post refers to finding a way to kill certain processes taking too much cpu resources not disk resources.

thks anyway