Bash Jobs and Processes Job handling

Help us to keep this website almost Ad Free! It takes only 10 seconds of your time:
> Step 1: Go view our video on YouTube: EF Core Bulk Insert
> Step 2: And Like the video. BONUS: You can also share it!

Example

Creating jobs

To create an job, just append a single & after the command:

$ sleep 10 &
[1] 20024

You can also make a running process a job by pressing Ctrl + z:

$ sleep 10
^Z
[1]+  Stopped                 sleep 10

Background and foreground a process

To bring the Process to the foreground, the command fg is used together with %

$ sleep 10 &
[1] 20024

$ fg %1
sleep 10

Now you can interact with the process. To bring it back to the background you can use the bg command. Due to the occupied terminal session, you need to stop the process first by pressing Ctrl + z.

$ sleep 10
^Z
[1]+  Stopped              sleep 10

$ bg %1
[1]+ sleep 10 &

Due to the laziness of some Programmers, all these commands also work with a single % if there is only one process, or for the first process in the list. For Example:

$ sleep 10 &
[1] 20024

$ fg %        # to bring a process to foreground 'fg %' is also working.
sleep 10

or just

$ %           # laziness knows no boundaries, '%' is also working.
sleep 10

Additionally, just typing fg or bg without any argument handles the last job:

$ sleep 20 &
$ sleep 10 &
$ fg
sleep 10
^C
$ fg
sleep 20

Killing running jobs

$ sleep 10 &
[1] 20024

$ kill %1
[1]+  Terminated              sleep 10

The sleep process runs in the background with process id (pid) 20024 and job number 1. In order to reference the process, you can use either the pid or the job number. If you use the job number, you must prefix it with %. The default kill signal sent by kill is SIGTERM, which allows the target process to exit gracefully.

Some common kill signals are shown below. To see a full list, run kill -l.

Signal nameSignal valueEffect
SIGHUP1Hangup
SIGINT2Interrupt from keyboard
SIGKILL9Kill signal
SIGTERM15Termination signal

Start and kill specific processes

Probably the easiest way of killing a running process is by selecting it through the process name as in the following example using pkill command as

pkill -f test.py

(or) a more fool-proof way using pgrep to search for the actual process-id

kill $(pgrep -f 'python test.py')

The same result can be obtained using grep over ps -ef | grep name_of_process then killing the process associated with the resulting pid (process id). Selecting a process using its name is convinient in a testing environment but can be really dangerous when the script is used in production: it is virtually impossible to be sure that the name will match the process you actually want to kill. In those cases, the following approach is actually much safe.

Start the script that will eventually killed with the following approach. Let's assume that the command you want to execute and eventually kill is python test.py.

#!/bin/bash

if [[ ! -e /tmp/test.py.pid ]]; then   # Check if the file already exists
    python test.py &                   #+and if so do not run another process.
    echo $! > /tmp/test.py.pid
else
    echo -n "ERROR: The process is already running with pid "
    cat /tmp/test.py.pid
    echo
fi

This will create a file in the /tmp directory containing the pid of the python test.py process. If the file already exists, we assume that the command is already running and the script return an error.

Then, when you want to kill it use the following script:

#!/bin/bash

if [[ -e /tmp/test.py.pid ]]; then   # If the file do not exists, then the
    kill `cat /tmp/test.py.pid`      #+the process is not running. Useless
    rm /tmp/test.py.pid              #+trying to kill it.
else
    echo "test.py is not running"
fi

that will kill exactly the process associated with your command, without relying on any volatile information (like the string used to run the command). Even in this case if the file does not exist, the script assume that you want to kill a non-running process.

This last example can be easily improved for running the same command multiple times (appending to the pid file instead of overwriting it, for example) and to manage cases where the process dies before being killed.



Got any Bash Question?