Help with simple script to find PID of process

thirdcoaster · October 30, 2012, 2:12pm

Hi everyone. I've been reading around and am a little bit overwhelmed, hoping to find a kind soul out there to hold my hand through writing my first script. This need has emerged at work and I haven't much experience writing shell scripts, but this is a problem we have with a production environment and I am tasked with providing the solution. After 3 days of staring and trying to Frankenstein together a solution, I'm no closer than I was when I started. I need some help, and so am kindly requesting it, with crossed fingers. I've put on my flame-retardant shield and am admitting I am a complete newbie off the bat. I'm at your mercy!

Here is the situation: we have a java process that keeps crashing. Normally in this situation we would monitor the existence of the pid file to tell us when it is available, but in this case the process is crashing due to a memory fault, which is leaving the pid file stranded. Therefore, it is necessary to periodically run a command, 'jps -l | grep cq' to see if the process is actually available, and interpret the result.

I have a few unique challenges in this case. First, I need to monitor on three different production boxes (all SunOS), but the directory where jps is kept is in two different locations. On one box it is:

/usr/jdk/jdk1.5.0_24/bin

and the other:

/usr/jdk/instances/jdk1.5.0/bin

Once I'm able to determine which location the command is to be run from, the command should be run, and if a line is returned then result is 0, if nothing is returned then 1.

It seems so simple but I'm lacking the know-how to put it all together.

I have bits and pieces but I'm sure there are large holes in the script:

if [-f $CMDPATH1\ $CMDPATH2
THEN
CMDPATH=$CMDPATH1
ELSE
CMDPATH=$CMDPATH2
fi
 
IF [ $PID == 0 ]
THEN
ERROR=1
ELSE
ERROR=0
FI
 
 
CMD="jps -l | grep cq"

So...help? If I left anything out please let me know - I'm very closely monitoring the thread. Thank you in advance.

Yoda · October 30, 2012, 2:59pm

Is this what you are trying to do?

CMDPATH1="/usr/jdk/jdk1.5.0_24/bin"
CMDPATH2="/usr/jdk/instances/jdk1.5.0/bin"

if [ -d $CMDPATH1 ] 
then
      # $CMDPATH1/cmd - Execute your command
elif [ -d $CMDPATH2 ]
then
      # $CMDPATH2/cmd - Execute your command
fi

thirdcoaster · October 30, 2012, 3:10pm

That looks so tidy and nice!

The second part is I guess, how to interpret the output and pass that into a message.

So, the output of the command, if everything is hunky dory, returns the following:

root@system>/usr/jdk/instances/jdk1.5.0/bin> jps -l | grep cq
11604 cq5-author-4502.jar

If not, nothing comes back. How do I make that input readable? This is such a basic question but I hope I'm in the right place - I need to "echo" this output, and in order to do so need to store the output as a variable, correct?

Thank you for your reply, I really appreciate you helping me get through this!

Yoda · October 30, 2012, 3:27pm

If you want to check that the process: cq5-author-4502.jar is running and do some action, then you can try:-

if [ `jps -l | grep cp| wc -l` -eq 0 ]
then
      echo "Process is not running"
      # Or do what you want to do here.
else
      echo "Process is running"
      # Ignore
fi

thirdcoaster · October 30, 2012, 4:09pm

Thank you!!!

thirdcoaster · November 5, 2012, 8:45am

#set variables
CMD="jps -l | grep cq | wc -l"
CMDPATH1="/usr/jdk/jdk1.5.0_24/bin/"
CMDPATH2="/usr/jdk/instances/jdk1.5.0/bin/"

if [ -d $CMDPATH1 ]
then
        $CMDPATH1/$CMD
elif [-d $CMDPATH2 ]
then
        $CMDPATH2/$CMD
fi

if [ 'jps -l | grep cq | wc -l' -eq 0 ]
then
        echo "Process is not running"
else
        echo "Process is running"
fi

exit 0

This is what I've got so far. I do get an "invalid argument count" when I run it:

root@server>/var/opt/OV/bin/instrumentation> ./cq_mon.sh cq_mon
invalid argument count
usage: jps [-help]
       jps [-q] [-mlvV] [<hostid>]

Definitions:
    <hostid>:      <hostname>[:<port>]
./cq_mon.sh[57]: jps -l | grep cq | wc -l: bad number
Process is running

Can anyone tell me what's happening here?

vgersh99 · November 5, 2012, 8:51am

change:

if [ 'jps -l | grep cq | wc -l' -eq 0 ]

to this:

if [ `jps -l | grep cq | wc -l` -eq 0 ]

pamu · November 5, 2012, 8:58am

OR..

if [[ ! $(jps -l | grep cq) ]]

RudiC · November 5, 2012, 9:27am

Not sure how far your SUN OS / shell supports this, but you might want to try

 jps -l | grep [c]q && echo "proc running" || echo "proc not running"

grep will leave an exit code: 0 if found, 1 if not found. && will execute the next statement, or even list, if found; || will do if not found. The [c] makes grep not find itself; and you may want to be way more specific in the process pattern that you grep for.

thirdcoaster · November 5, 2012, 3:22pm

Unfortunately that didn't work either. I made the suggested change and get the following:

root@quartz>/var/opt/OV/bin/instrumentation> ./cq_mon.sh cq_mon
invalid argument count
usage: jps [-help]
       jps [-q] [-mlvV] [<hostid>]

Definitions:
    <hostid>:      <hostname>[:<port>]
./cq_mon.sh[57]: jps:  not found
Process is not running

My concern here is that the process is in fact running...previously it would at least echo the right message. My question is now whether something is wrong with that last 'if' line:

if [ `jps -l | grep cq | wc -l` -eq 0 ]

is being run from the right dir, since the error indicates that jps is not found. :wall:

---------- Post updated at 03:22 PM ---------- Previous update was at 03:10 PM ----------

rudic:

Not sure how far your SUN OS / shell supports this, but you might want to try
 jps -l | grep [c]q && echo "proc running" || echo "proc not running"
grep will leave an exit code: 0 if found, 1 if not found. && will execute the next statement, or even list, if found; || will do if not found. The [c] makes grep not find itself; and you may want to be way more specific in the process pattern that you grep for.

I appreciate you explaining what each piece does, THANK YOU! I am bumping up against the same problem though, the command jps isn't being found. Somehow I think I need to pass the location of jps from the first part of the script into the bottom.

root@server>/var/opt/OV/bin/instrumentation> ./cq_mon.sh cq_mon
invalid argument count
usage: jps [-help]
       jps [-q] [-mlvV] [<hostid>]

Definitions:
    <hostid>:      <hostname>[:<port>]
./cq_mon.sh[57]: jps:  not found
proc not running

Yoda · November 5, 2012, 3:25pm

Can you re-try using the absolute path of jps in your if statement?

thirdcoaster · November 5, 2012, 3:30pm

Yes. I did that and got a better result - the pid is returned and the correct message comes through: proc running. However, there is still an invalid argument error, and I am still stuck with the problem that this script is to be run on three different systems, two of which have a different absolute path for jps. Result follows:

root@server>/var/opt/OV/bin/instrumentation> ./cq_mon.sh cq_mon
invalid argument count
usage: jps [-help]
       jps [-q] [-mlvV] [<hostid>]

Definitions:
    <hostid>:      <hostname>[:<port>]
350 cq5-author-4502.jar
proc running

RudiC · November 6, 2012, 3:10am

thirdcoaster:

root@quartz>/var/opt/OV/bin/instrumentation> ./cq_mon.sh cq_mon
invalid argument count
usage: jps [-help]
   jps [-q] [-mlvV] [<hostid>]

Definitions:
   <hostid>:      <hostname>[:<port>]

Are you sure you want to run jps ? From that errror msg I infer it does not report process states etc. but info about remote hosts. Why don't you use ps ?

thirdcoaster · November 6, 2012, 8:10am

It is a requirement of my business partner that I use jps. jps will return the information I need, which is basically, is there a cq process running or not. I have confirmed this many times.

The only outstanding issue is that I'm still getting the error, which I believe is because in the last call of jps the directory isn't getting passed correctly. That's what I need help with.

Thanks.