183

How do I run a program in the background of a shell, with the ability to close the shell while leaving the program running? Lets say my UI is having problems or for some reason, I need to boot up a program from the terminal window, say, nm-applet:

nm-applet

When it's started, it occupies the foreground of the terminal window.

Is there any simple way to run the program in the background without needing to leave the terminal open or have it occupy the whole terminal?

On that note, I did find a way to run programs from the terminal and have it allow for other inputs, by appending an ampersand (&) to the command as such:

nm-applet &

But this isn't much use as any processes started in the terminal are killed once the terminal is closed.                                                                        

Nicolas Barbulesco
  • 103
OVERTONE
  • 2,307

10 Answers10

221

I've recently come to like setsid. It starts off looking like you're just running something from the terminal but you can disconnect (close the terminal) and it just keeps going.

This is because the command actually forks out and while the input comes through to the current terminal, it's owned by a completely different parent (that remains alive after you close the terminal).

An example:

setsid gnome-calculator

I'm also quite partial to disown which can be used to separate a process from the current tree. You use it in conjunction with the backgrounding ampersand:

gnome-calculator & disown

I also just learnt about spawning subshells with parenthesis. This simple method works:

(gnome-calculator &)

And of course there's nohup as you mentioned. I'm not wild about nohup because it has a tendency to write to ~/nohup.out without me asking it to. If you rely on that, it might be for you.

nohup gnome-calculator

And for the longer-term processes, there are things like screen and other virtual terminal-muxers that keep sessions alive between connections. These probably don't really apply to you because you just want temporary access to the terminal output, but if you wanted to go back some time later and view the latest terminal activity, screen would probably be your best choice.

The internet is full of screen tutorials but here's a simple quick-start:

Artur Meinild
  • 26,018
Oli
  • 293,335
  • Oli, in the case of DISPLAY=:0 unity --replace which of the above would you use and why? – nutty about natty Feb 07 '13 at 12:40
  • Also, in my command line "alarm clock" sleep 10s && cvlc '/home/omm.ogg' I've tried to add & disown, but after the 10s when (c)VLC is launched, the process takes over the command line again. Is there an easy way to prevent that from happening? – nutty about natty Mar 12 '13 at 14:44
  • 4
    @nuttyaboutnatty wrap the whole lot in its own shell session: sh -c "sleep 10s && cvlc '/home/omm.ogg'" & disown. That's pretty much my solution for everything to make sure it forks out properly. – Oli Mar 12 '13 at 14:56
  • @nuttyaboutnatty In answer to the older query, whatever works :) When I bug out to a TTY for relaunching a compositor (where I've lost Alt+F2), I'll run it from the TTY and then run it again from Alt+F2's run dialogue, effectively ending the TTY-run version... So it's not an issue. If you're having problems with output lingering on, you need to look at redirection which disowning et al won't affect that much. – Oli Mar 12 '13 at 15:01
  • you mean, run DISPLAY=:0 unity --replace first from a tty1, and then again from tty7? Why would you want to run it twice? In my (limited) experience, I'd run it once in tty1, then switch to tty7, back to tty1, back to 7 (without doing anything, only observing the successive changes as I switch to and fro), and finally, when tty7 has sorted itself out (after maybe a minute or so?) I'd one last time switch to tty1, hit ctrl c, exit, and switch back to tty7 (and reboot, depending on my mood and the constellation of the skies)... – nutty about natty Mar 12 '13 at 15:10
  • 1
    @nuttyaboutnatty That's what I mean but you wouldn't end up with two instances running because --replace does what it sounds like and actually ends the old instance. Running it the second time from within the xsession would end the copy bound to TTYn and would let you use that in the future if you needed it. There is probably a better way of doing it but just what comes to my mind each time I need it. – Oli Mar 12 '13 at 15:25
  • 1
    Perfect, the one that worked for me was the subshell '(gnome-calculator &)'. I use it to launch a mono console app in the background. With the other techniques, the mono app would crash instantly. – Nikolaos Georgiou Oct 28 '13 at 20:55
  • 6
    Note to future users: creation of nohup.out can be suppressed with redirecting both stdout and sterr to /dev/null like so: nohup firefox &> /dev/null & – Sergiy Kolodyazhnyy Aug 07 '16 at 18:13
74

Here's the two ways I'd go with. Firstly, not running it from a terminal; hit Alt+F2 to open the run dialog, and run it from there (without &).

From a terminal, run

nm-applet &

But do NOT close the terminal yourself. That is, do not hit the X-button to close, and do not use File -> Exit from its menubar. If you close the terminal that way, it will send a HUP (Hang UP) signal to the bash running within, which in turn will send the HUP signal to all its children (which is why nohup works in this case).

Instead, exit the shell by running exit or hitting Ctrl+D. bash will then disown its children, then exit, leaving the background processes still running. And when bash exits, the terminal has lost its child process, so it will close too.

Doing it all at once:

nm-applet & exit
geirha
  • 46,101
  • 1
    That's amazing! Never knew that and you seem to be the only person to mention this. Definitly that's my favorite approach when in a Desktop Environment. Too bad I can't upvote more ;p – 7hi4g0 Jan 24 '15 at 06:01
  • I tried this with a different application and it failed: matlab & exit Running as two separate commands worked well: matlab & exit – MattKelly Nov 10 '16 at 20:00
  • I was wondering why OP said closing the terminal caused his background programs to crash. I've never tried closing a terminal by the "x" button on the window. – user1717828 Mar 29 '17 at 12:51
  • the x button is kill it doesn't cleanly exit, the exit command does. you could remap the x button to a exit command instead... – Tim Feb 02 '18 at 07:35
19

As you pointed out, you can run 

nohup nm-applet &

to ignore the end signal when closing the terminal. No problem with that.

desgua
  • 32,917
  • Any other alternatives? Just for knowings sake, not for anything else – OVERTONE Feb 22 '12 at 00:38
  • At wikipedia ( http://en.wikipedia.org/wiki/Nohup ) there is a suggestion to use echo command | at now which I couldn't get it to work. – desgua Feb 22 '12 at 00:45
  • 2
    I wonder how it works from a the UI when you double click an icon or program. – OVERTONE Feb 23 '12 at 10:22
  • exec provides an alternative to nohup <command> & for running a process in the background and returning current shell prompt. – jsstuball Jul 30 '19 at 09:16
18

One thing that many other answers are missing is how to detach a running process that currently blocks the shell. In most terminals and shells, Ctrl+Z will halt the running process and bring you back to an input prompt. Then, you can issue

bg

to send the running process into the background. Issue

fg

instead to put the running process back into the foreground.

EDIT: More detail in this answer I discovered later.

Community
  • 1
krlmlr
  • 3,337
10

Use （exec PROGRAM &> /dev/null & ） to allow PID of subshell to be taken over by PROGRAM. I've tested this approach multiple times with several different programs. Closing the original terminal has no affect on the newly-spawned program

Small demo:

$ # this is before running
$ (exec firefox &> /dev/null &)
$ # and look, we still in side the terminal and can continue working
Sergiy Kolodyazhnyy
  • 105,154
  • 20
  • 279
  • 497
2

I can recommend the byobu terminal. You can easily detach your process by pressing the F6 key.

speter
  • 471
2

Although there are good answers above, I would like to give my 2 cents on how I use MATLAB in background.

sudo -b matlab

The -b tag of sudo can run applications in background. You can also close the terminal after executing this command.

Probably, unrelated but there is wonderful website that you can use to explain shell commands. http://explainshell.com/explain/8/sudo

user3342981
  • 21
  • 2
1

The modern and easy to use approach that allows managing multiple processes and has a nice terminal UI is hapless utility.

Install with pip install hapless (or python3 -m pip install hapless) and just run

$ hap run my-command  # e.g. hap run python my_long_running_script.py
$ hap status  # check all the launched processes

See docs for more info.

ui

NOTE: I'm the author of this package.

Most Wanted
  • 151
0

In case KDE is used, you can also use kstart, which will start your program detached from the terminal. It also makes sure that the KDE environment is correctly setup for the command. (See kstart.cpp source code for reference. As you see from there, it uses KProcess::startDetached, and KProcess is derived from QProcess, and startDetached starts a new process, and detaches from it.)

Similar is also kde-open or xdg-open or gnome-open.

Albert
  • 2,577
  • As this stands, it looks almost more like a comment than an answer. Please [edit] to expand. Don't just give a one-liner; explain why you right, ideally with citations. Answers without explanation are subject to removal. – anonymous2 Mar 05 '19 at 13:49
0

I dont know its the right way but i just start another session while leaving the previous one allone. For example i ran a simple web server on my raspberry. the web.py one. then i start a new session while leaving it alone. thats it. it also is wuite useful since you are still updated even though you are workin on the other session.

leo
  • 1