[SOLVED] Determining if a process is running

[dwhitney67]

Is there an easier way to determine if a process is running, other than using the "ps" (process status) command?

A co-worker insists the following script is inefficient and not because of the "sleep" call, but because of the repeated calls to "ps". I agree with him, thus I am looking for a way to perform the "ps" only once, and use this result through each iteration of the loop.


#!/bin/sh

isProcessRunning()
{
# $1 is process-name to verify if running
#
ps -ef | cut -b 49- | grep -e ^.*/$1$ >/dev/null 2>/dev/null
retval=$?
}


# Main program
#
while true
do
# DATABASE
isProcessRunning database
if [ $retval -ne 0 ]
then
/usr/bin/database > /dev/null 2> /dev/null &
fi

# ARCHIVER
#isProcessRunning archiver
#if [ $retval -ne 0 ]
#then
# /usr/bin/archiver > /dev/null 2> /dev/null &
#fi

# HISTOGRAM
isProcessRunning histogram
if [ $retval -ne 0 ]
then
/usr/bin/histogram > /dev/null 2> /dev/null &
fi

# HARDWARED
isProcessRunning hardwared
if [ $retval -ne 0 ]
then
/usr/bin/hardwared > /dev/null 2> /dev/null &
fi

sleep 2
done

[scorp123]

You may want to look into the pidof command ... It returns the PID of a process. Let's suppose I wanted to check if Firefox is running:
$ pidof firefox-bin
13549 ... The number there is the PID of my firefox instance. If nothing is returned then the program in question isn't running. This can be also incorporated into a script, e.g. this command:
echo $? ... returns the exit code of the last command that was run. So my example with "pidof firefox-bin" gives me a return code of "0" which means "all OK" whereas e.g. "pidof blahblahcatchme" will give me a return code of "1" which means "not OK".

And of course you could put this into a nice "if ... then" loop, e.g.
if [ "$?" -ne "0" ]; then
echo "Sorry, program xyz is not running it seems!"
exit 1
fi

So I suppose your script up there could be shortened and made to look more elegant?

[dwhitney67]

Thanks for the tidbit. I will look into whether my non-Ubuntu system supports this. This command is definitely much more efficient than doing the "ps", followed by a "cut", and then a "grep".

[scorp123]

I will look into whether my non-Ubuntu system supports this. This command is definitely much more efficient than doing the "ps", followed by a "cut", and then a "grep". The "pidof" command is pretty much standard on Linux .... but not on commercial UNIX, e.g. Solaris, HP-UX, and so on. So if your script is running on one of those OS'es you'd have to do it the way you did it now with all those tools.

[tyoc]

IIRC, you can do some like kill -0 or some like that.

[scorp123]

IIRC, you can do some like kill -0 or some like that. Taken from the "kill" manual:
SIGNALS
The signals listed below may be available for use with kill.
When known constant, numbers and default behavior are shown.

Name Num Action Description
0 0 n/a exit code indicates if a signal may be sent
....


So this would check if a task is still responding (= accepting signals via "kill") .... that's kinda not exactly the same.

[PaulFr]

Just my 0.02: In POSIX systems,
ps -eo comm|grep <command_name>is reasonable (i.e. look at the -o option to ps). On Solaris, pgrep has some useful arguments.

[dwhitney67]

Everyone, thank you for your replies. I think I will rely on scorp123's suggestion and use the pidof command. The system I am using, CLFS (cross-compiled linux from scratch) has support for this command.

My script should end up looking something like:


#!/bin/sh

while true
do
# DATABASE
pidof database >& /dev/null
if [ $? -ne 0 ]
then
/usr/bin/database >& /dev/null &
fi

# ARCHIVER
#pidof archiver >& /dev/null
#if [ $? -ne 0 ]
#then
# /usr/bin/archiver >& /dev/null &
#fi

# HISTOGRAM
pidof histogram >& /dev/null
if [ $? -ne 0 ]
then
/some/path/not/usr/bin/histogram >& /dev/null &
fi

# HARDWARED
pidof hardwared >& /dev/null
if [ $? -ne 0 ]
then
/usr/bin/hardwared >& /dev/null &
fi

sleep 2
done

[geirha]

Instead of

pidof database >& /dev/null
if [ $? -ne 0 ]
You can simply do

if ! pidof database >& /dev/null
And since you only run one command in each if, you can make it quite compact with

pidof database >& /dev/null || /usr/bin/database >& /dev/null & The command after || will only run if the first command fails (returns non-zero)

[scorp123]

pidof database >& /dev/null || /usr/bin/database >& /dev/null & The command after || will only run if the first command fails (returns non-zero) ... Which is the opposite of
command1 && command2 So command2 will only execute if command1 before it was successful.

(Just for the sake of completeness)