UNIX for Dummies Questions & Answers

Hi

I'm writing a script which will be run by cron every X minutes.

I don't want cron to run my script again if the previous one has not yet finished.

When the script first runs, I had the idea to store the Process ID in a file. When cron tries to run the script again, I would check the file to see if the previous process ID still existed.

My dilemma is the following:

1. How likely is it that a totally seperate process may have taken the original PID in the file, hence I will be checking the wrong process.

2. Is there any better way to see if my script has finished?

Thanks
Helen

here is something someone posted a while back, I couldn't find the thread but had saved the reply:

Easiest way is:

ps -ef | grep "[l]d_data" || ld_data


That will run the program "ld_data" if and only if it is not found already running. Placing the first character within square brackets protects the "grep" command line from being found and triggering a false positive.

Nice little tool, although in this case it doesn't seem to work.
For some reason the scriptname is not displayed just '-sh'

ie if I do a ps -ef and grep for the process ID I get

helen 25452 16598 255 16:34:53 pts/tg 1:57 -sh

I thought the script name should appear insead of -sh.

Any ideas?

Thanks
Helen

In a korn shell you can do the following:

Add to your script a lock condition.

Example:

while [ `ps -ef |grep your_script_name |grep -v grep |wc -l ` -gt 1 ]
do
echo "waiting to the end of the previous instance"
sleep 10
done

#your code
.....


Regards. Hugo.

There is no perfect solution to this problem. Someone else on the system might be running an identically named script. And pids do recycle.

Here is a quick script that illustrates my solution:

Code:

#! /usr/bin/ksh

if [[ -f lockout.pid ]] ; then
        if kill -0 $(cat lockout.pid) ; then
                echo still running
                exit 0
        fi
fi
echo $$ > lockout.pid
trap "rm lockout.pid" 0
sleep 30
exit

Is the script is functioning properly, it will create and delete the lockout.pid file correctly. If the lockout.pid file exists, then the process should also exist.

If the script is abruptly killed and cannot remove the lockout.pid file then the next instance of the script will detect it. The new instance will try to send signal zero to the old pid. If the script is not running as root, it will not be able to signal another user's process.

If the previous instance of the script finished normally, the new instance of the script cannot be fooled. If the previous instance of the script is still running, the new instance of the script cannot be fooled. These are the normal conditions we expect to find.

So to fool the script, first it must abort, second the pid must have recycled and the pid must now be in use by a new, unrelated process, and third, the new process be belong to the same user that is running the script. The third condition is lifted if root runs the script.

Instead of searching for a process number you could have your script create a message file when the script is finished running.
EX:

echo 'finished' > /tmp/finished

If you do this you will have to edit the begining of the script also. You will have to search for the created file and if it doesn't exist stop the process from running again if it does remove it and continue running the script.

*NOTE: it is important to remove the file BEFORE running the script again.

General outline:

search for /tmp/finished

if (not there)
then
quit process
else
rm /tmp/finished and run process
fi

echo 'finished' > /tmp/finished