I have a script (run.sh) which calls my application script. At a given time, I may have multiple run.sh running.

The general format of the run.sh script is,

# Running DALTON JOB: Helix
    echo "-----------------------------------------------"
   export DALTON_TMPDIR=/mnt/raid0/scratch
    export OMP_NUM_THREADS=6
    source /opt/intel/compilers_and_libraries_2017.0.098/linux/bin/compilervars.sh intel64
    source /opt/intel/mkl/bin/mklvars.sh intel64
    echo "//-------process started----------------------------//"

./application.sh  -mb 14550 input.mol output.out

echo "//-------process finished----------------------------//"

Is it possible to get the PID of the application.sh inside the run.sh script. (I found that $$ gives the PID of the script itself.)

Also, I noticed that the PID of the application is always larger numerically than parent script but maybe its coincidence.

  • 1
    does ./application.sh get run in the background? otherwise, it will have exited by the time run.sh could do anything with the PID. – Jeff Schaller Jun 8 '17 at 14:14
  • application.sh runs in the background ( for ~3-4 hours) as a host process of run.sh. – ankit7540 Jun 8 '17 at 14:15
  • 1
    Aside: the PID is assigned in increasing numerical order, so in general parents have a lower PID number than children. However, there is a point where the PID wrap. E.g., the maximum PID is 32768, so it is possible for a child process to have a lower PID number than a parent. (note: 32768 is an example and not necessarily the max value on a given system). – KevinO Jun 8 '17 at 14:22
  • 1
    eliding the actual call to application.sh with "parameters" muddies the waters. show that it goes into the background, then I think this Q has a usable answer. – Jeff Schaller Jun 8 '17 at 14:25
  • 1
    I'm confused. You do not put it in the background; it runs for 3-4 hours, and then you get the "echo process finished"? The application.sh has exited by that point. What PID do you want, and where? – Jeff Schaller Jun 8 '17 at 14:33

If you want to see the PID of application.sh while it is running, then I would suggest explicitly putting it into the background, capturing the PID, then waiting for it to exit:

# ...
./application.sh  -mb 14550 input.mol output.out &
echo "The application pid is: $app_pid"
wait "$app_pid"
# ...
  • I would second this. '$!' captures the pid of the last executed (backgrounded) command which is the command in question. – LJKims Jun 8 '17 at 14:52
  • I need a clarification. wait "$app_pid" keeps the run.sh waiting till application.sh finishes right ?? – ankit7540 Jun 8 '17 at 14:52
  • 2
    That's correct. – LJKims Jun 8 '17 at 14:53

I believe you may want a combination of something like pstree -p and ps axjf with some additional parsing.

Notice how the PID of httpd is 30469, and every process that is a child of httpd has a PPID (parent process ID) of 30469. The grandchildren of httpd will have the PPID of their parent process, which will have the PPID of the httpd process.

I didn't post the full output of either as they are rather large. Here's an example output from each:

user@host$ ps -axjf
1     30469 30469 30469 ?           -1 Ss       0   0:46 /usr/sbin/httpd
30469 22410 22410 22410 ?           -1 Ssl      0   0:00  \_ PassengerWatchdog
22410 22413 22413 22410 ?           -1 Sl       0  23:06  |   \_ PassengerHelperAgent
22410 22418 22418 22410 ?           -1 Sl      99   0:01  |   \_ PassengerLoggingAgent
30469 22442 30469 30469 ?           -1 Sl      48   7:55  \_ (wsgi:pulp)
30469 22443 30469 30469 ?           -1 S       48   1:48  \_ /usr/sbin/httpd
30469 22445 30469 30469 ?           -1 S       48   1:55  \_ /usr/sbin/httpd
30469 22447 30469 30469 ?           -1 S       48   1:54  \_ /usr/sbin/httpd

user@host$ pstree -p
        │              │                        │                        ├─{PassengerHelpe}(22416)
        │              │                        │                        ├─{PassengerHelpe}(22417)
        │              │                        │                        ├─{PassengerHelpe}(22420)
        │              │                        │                        ├─{PassengerHelpe}(22422)
        │              │                        │                        ├─{PassengerHelpe}(22423)
        │              │                        │                        ├─{PassengerHelpe}(29342)

Note if you know the parent process tree you can run pstree -p <pid>.


To save the process ID of a utility, start it as an asynchronous process and save its PID in a variable, or use $! directly (it's the PID of the most recently started background process):


export DALTON_TMPDIR=/mnt/raid0/scratch
source /opt/intel/compilers_and_libraries_2017.0.098/linux/bin/compilervars.sh intel64
source /opt/intel/mkl/bin/mklvars.sh intel64

printf 'Started at "%s"\n' "$(date)"

./application.sh -mb 14550 input.mol output.out &


printf 'Application PID is %d\n' "$app_pid"

wait "$app_pid"

printf 'Ended at "%s"\n' "$(date)"
printf 'Run time was approximately %d minutes (%d seconds)\n' "$(( SECONDS/60 ))" "$SECONDS"

The PID of the application is $app_pid.

You say you have noticed that the PIDs of new processes are always larger than those of previous processes. This is nothing that you may rely on for two reasons:

  1. There will be a PID allocation wrap-around once the maximum allowed PID has been allocated to a process, and old PIDs will start to be reused. The next PID will be smaller, not larger, than the previous PID when the wrap-around happens (and new PIDs after that will skip numbers that are still in use as PIDs by running processes).
  2. Some systems, like OpenBSD, use randomized PID allocation:

    $ for i in {1..5}; do bash -c 'echo $$'; done

As a side note, you may use $PWD rather than $(pwd) to get the current working directory.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.