AIX

Hi,

We have scheduled few jobs/scripts in cron and it runs on daily basis at a specified time. We have a cron jobs has various parameters for main scripts like

/home/test/mainscript.ksh main_param

when the above jobs triggers, mainscript.ksh has below code and it refers

main_param="param1 param2 param3 param4"

the script1.ksh triggeres one by one for these params like script1.ksh param1 (once it finishes its execution), then it triggers script1.ksh param2

So, consider script1.ksh param1 is still running till next day the same param1 triggers (as said cron jobs runs on daily basis), it overlaps the logs and makes more usage in the database/server.

In order to avoid that, I have included the below code

Code:

check=`ps -ef | grep-w "script1.ksh param1"| grep -v "grep" | wc -l`
if [ $check -gt 1 ]; then
echo "Already process running"
fi

It was running perfect in Sun solaris box, but in AIX, it started misbehaving like when the next param runs on the same day, it says that älready process is running"

script1.ksh param1 -- after it completion, it says that script1.ksh param2 is running (more than 1 instance condition) but there is no chance of param2 is running currently.

I have checked the condition as $check -gt 2, but it throws the message rarely that "Already process running"

Please help me to identify the correct process and throw this message.

your timely help is much appreciated as I m in the neck moment.

You avoid the ps | grep | grep | awk | sed | cut | wc -l | kitchen | sink junk as well as implementation-dependent ps difference problems, by using PID files.

At the beginning of your scripts:

Code:

if [ -f /tmp/myscriptlock ]
then
        read PID < /tmp/myscriptlock
        if ps $PID >/dev/null
        then
                echo "Already running" >&2
                exit
        else
                echo "PID file but no process, did it quit unexpectedly?" >&2
        fi
fi

echo $$ > /tmp/myscriptlock
trap "rm -f /tmp/myscriptlock" EXIT

# rest of your script

The trap .......... EXIT is very useful, but make sure that your code does not exec another program or reset the traps anywhere or you will not clean up the file. I've fallen over that one.

Also be sure that you don't have something that cleans up /tmp as I have fallen over this before when someone thought it would be a good idea to remove temporary files at 2am every night. It gets worse when the schedule runs late, but it's all so badly written that we need something to clean up every now and then so without a huge re-write of the schedule we're stuck. Can you see I'm experienced in it all going wrong?


This is still a much better idea than just counting processes, as Corona688 says. Of course, you need to ensure that /tmp/myscriptlock is unique in every script. Perhaps you could base it on the script actually running with something like:-

Code:

lockfile="${0##*/}.lock"

I hope that this helps,

Robin
Liverpool/Blackburn
UK

The traditional place for PID files is actually in /var/run for that reason, but user scripts usually can't create files there, just daemons.