Script Help - killing a job if its more then 2 hours old

[Larshg]

Hi

I have a job that starts in the crontab every minute – it only runs for 2 sec. – but in some rare cases it gets stuck. – this means that the process never dies, and this in turn means that it can’t start next time – and that means that the job stops working.

The only think to do is to kill the process, so that it can start up again.

I need to make a script that does this – the only thing the script should do is to kill the process IF it has been running for more then 2 hours


Hope some one can help

 

[gregor weertman]

You can do it difficult or the easy way.
“ps” is the first way.

The script you start with the crontab must make a lock-file.

Now write the process id of the script in yourprogram.lck
echo $$ > /tmp/yourprogram.lck

The next time you run the program first check if yourprogram.lck exists.
If so do:
echo “-“ >> yourprogram.lck ## add a line
if [ $(cat /tmp/yourprogram.lck |wc –l) –lt 120 ] ;then
kill `head -1 yourprogram.lck`
else
exit
Fi
(Your program runs every minute.)

to sum up.
First check if the program is running.
When not: echo $$ > /tmp/yourprogram.lck
When yes:
If [ $(cat /tmp/yourprogram.lck |wc –l) –gt 120 ] ;then
Kill -9 `head -1 yourprogram.lck`
else
exit
fi
and
echo $$ > /tmp/yourprogram.lck
remove /tmp/yourprogram.lck when finished

When it is a binary program you run, run it from the script like:
Binary-program&
echo $! > /tmp/yourprogram.lck to capture the process id of the binary.
Now you can not remove yourprogram.lck.
First check if the program is running by:
if [ -n “$(ps –ef|grep $(head -1 yourprogram.lck)|grep –v grep) ]
then
echo …



Success


Gregor.Weertman@mailcity.com

 

[Larshg]

Hmmm I not sure I fully understand your script - but I get the genneral ideer - to make a log file and check it thats if the process is still there in 2 houers.

But wouldn't it be posible to make the script check the systime in 'ps' command - all it needs to do is check if the time in the 'ps' is more then 2 hours old! (add to hours to the time and check with 'date' and kill it if it's not equal og less the the time it was startet)

 

[gregor weertman]

You are right, but it is a drag.
You have to calculate with the time.

This is more work then counting the number of lines in the logfile.

When you would give me more detail,
I could give you more detailed help.

Regards Gregor.Weertman@mailcity.com