Shell script to restart service automatically

[Cahyo Tri Nugroho]

Dear,

Is there anybody have shell script that can check running services and automatically start the service if it goes down?
TIA.

Regards,
Cahyo

[Steven E. Protter]

Shalom Cahyo,

Here is an example

export UNIX95=1
smproc=$(ps -C sendmail | wc -l)

if [ $smproc -le 0 ]
then
/sbin/init.d/sendmail start
fi

You have to write it yourself for the service you wish to monitor.

SEP

[Peter Godron]

Cahyo,
assuming you want to not test once, schedule Steven's script via cron to run at specified intervals. See man cron and man crontab

And/or see Steven's loop solution at:
http://forums1.itrc.hp.com/service/forums/bizsupport/questionanswer.do?threadId=1024003

[spex]

Steven,

I believe "if [ $smproc -le 0 ]" should be "if [ $smproc -le 1 ]", as 'ps' also returns column headings:

# UNIX95= ps -C foo123xyz | wc -l
1

Cahyo,

Another (less elegant) option:

# while :; do sleep 10; /sbin/init.d/your_daemon start; done > /dev/null 2>&1 &

'nohup' will allow it to persist after your session terminates.

PCS

[Yogeeraj_1]

hi cahyo,

one simple and clean solution would be to use cron to run the check script at regular interval.

e.g. Cron entry that will check your service every 15 minutes.

15 * * * * /users/scripts/check_service.sh 1>/users/scripts/logfiles/output-checkscript.crn 2>/users/scripts/logfiles/error-checkscript.crn
#
#*******************************************************************************
# min|hour |day |month|day |script
# | |of mo| |of wk|
#----|-----|-----|-----|-----|--------------------------------------------------
#*******************************************************************************
#*******************************************************************************
# END OF TABLE day0->Sunday day6->Saturday
#*******************************************************************************

hope this helps!

kind regards
yogeeraj

[Bill Hassell · Accepted solution]

A couple of things about using ps to restart a specific process:

> export UNIX95=1
> smproc=$(ps -C sendmail | wc -l)
> if [ $smproc -le 0 ]
> then
> /sbin/init.d/sendmail start
> fi

It's best never to export the UNIX95 variable, even in a script. UNIX95 enables more options in ps but also affects several other commands and libraries. Use UNIX95 as a temporary variable on the command line as in:

UNIX95=1 ps -C sendmail

Now to eliminate ps headers, use the -o option as in:

UNIX95=1 ps -C sendmail -o pid= -o args=

This eliminates the header line and makes the count correct. So the script would look like this:

MYPROC=sendmail
COUNT=$(UNIX95=1 ps -C $MYPROC -o pid= -o args= | wc -l)
if [ $COUNT -lt 1 ]
then
/sbin/init.d/$MYPROC start
fi

Now the above script is only an example. For instance, if a process is unexpectedly not running, there may be some cleanup tasks that must be accomplished first. With any automated 'fix' for a broken program, you need to investigate why the process is stopping unexpectedly and eliminate the problem rather than kickstarting it from cron.

[Cahyo Tri Nugroho]

Dear all,
Thank you for your help :)

Rgds,
Cahyo