Write a PID file manually from Bash script

Question

How could I write a Bash-script that runs a long running program and stores the programs process id in a separate file?

I want something like

#!/bin/bash
exec long_running_tool
echo `ps af |grep "long_running_tool" |awk '$5 == "long_running_tool" {print $1}'` > pid_file

However doing exactly this would execute the ps after tool has finished.

Is there a way to get the process id of the process created?

Your sample code won't work because exec doesn't fork, so the shell process is replaced by the long_running_tool process. Consequently, exec never returns and the following line is never reached. See help exec, execve(2). — Etienne Dechamps, Jan 27 '11 at 10:04

Jeremy Sturdivant · Accepted Answer · 2011-01-27 10:46:53Z

12

You can easily run the process in the background with "&", then get the PID of the background process using "$!"

#!/bin/bash
long_running_tool &
echo $! > pid_file

Then, optionally, wait $! if you want the shell to block until the process completes execution.

answered Jan 27 '11 at 9:40

Jeremy Sturdivant

2,09813 silver badges11 bronze badges

1

UPDATE: when I use the fg I get 'fg: 21: job (null) not created under job control' and the script does not block. Sorry. – DerMike Jan 27 '11 at 10:23
"wait $!" should have a similar effect to fg that I can confirm does in fact, work in a script. I should probably have checked that before posting. >.> my bad. – Jeremy Sturdivant Jan 27 '11 at 10:29

add a comment |

grawity · Accepted Answer · 2011-01-27 15:26:23Z

9

#!/bin/bash
echo $$ > fooapp.pid
exec fooapp

As mentioned earlier, exec replaces bash with the specified program, retaining the PID.

answered Jan 27 '11 at 15:26

grawity

277k43 gold badges569 silver badges648 bronze badges

add a comment |

2 Answers 2