Add process to be run at boot

I have a server program which I would like to add to my system boot. I tried to add it to '/etc/init.d/' and quickly found that this was not what I wanted (Thank you for a command line interface, haha). Where do I want to place this executable to be booted?

You need to learn about runlevels and runlevel scripts, which for many/most/all distros (for sure redhat & debian) are in /etc/rc?.d.

Those are shell scripts run by init (process #1); you can just include a command-line invocation in the appropriate place -- perhaps rc.local, which may be in /etc/init.d. Don't put an actual program in that directory, just add a line to a script.

After reading on run-level scripts is it a run-level script that I want to run this as or is there another location where I want to put this server process? I may just be understanding you incorrectly

Originally Posted by NuNn

After reading on run-level scripts is it a run-level script that I want to run this as or is there another location where I want to put this server process?

You could make up a runlevel script:

Code:

!#/bin/bash

/home/user/C/myserver

give it a name, place it with the other scripts (probably in either /etc/rc.d or /etc/init.d), make sure to "chmod 755 myscript" and then place a softlink in the appropriate runlevel directory:

ln -s /etc/init.d/myscript /etc/rc5.d/S50myscript

for example. The link must be named "S" (for start) with a number (the priority) appended to the name of the script which exists in /etc/init.d (or /etc/rc.d, depending on the distro).

Or, like I said, just add a line to an existing script (rc.local is a good one to use for this stuff).

Originally Posted by MK27

You could make up a runlevel script:

Code:

!#/bin/bash

/home/user/C/myserver

A typical runlevel script takes a parameter, which is usually "start" or "stop" (but might also support "reload", "status", "restart" and a few others, but "start" and "stop" are usually required). So a good runlevel script would look like:

Code:

#!/bin/bash

case $1 in
    start)
        do_startup_stuff
        ;;
    stop)
        do_stopping_stuff
        ;;
esac

The part about linking it into /etc/rc.*/... is perfectly correct though.

EDIT: Some weird Linux distros don't use this system, but most of them do.

EDIT EDIT: The other problem is that runlevel scripts start the processes as root. Sometimes that is not what you want, and you have to play tricks.

Thank you MK/brewbuck that helps clear it up a bit for me. Another question is does it matter if I choose init.d or rc.d and if it is rc.d that I want I have a number of directories ranging from rc[1-5].d and one number rcS.d.

Originally Posted by NuNn

Thank you MK/brewbuck that helps clear it up a bit for me. Another question is does it matter if I choose init.d or rc.d and if it is rc.d that I want I have a number of directories ranging from rc[1-5].d and one number rcS.d.

The standard is to place the script itself into /etc/init.d, and then symlink it into the appropriate rc.* directory.

rc.3 is for normal system startup, and rc.6/rc.0 is for normal system shutdown. You probably want to link it into both of those locations. The "S50service" syntax is how you control the exact order in which services start up. Smaller numbers start/stop earlier, larger numbers do it later. If it doesn't matter (i.e. your service depends on no other service) then just pick a number.

Originally Posted by brewbuck

A typical runlevel script takes a parameter, which is usually "start" or "stop"

True, but the simple method (no start or stop) will work.

Just depends how interested you are in writing runlevel scripts.

Originally Posted by MK27

True, but the simple method (no start or stop) will work.

Just depends how interested you are in writing runlevel scripts.

Well, if you ignore the $1 parameter, then you need to write two scripts -- one for startup, one for shutdown. You could certainly do it that way, but maintaining two scripts is worse than maintaining one.

If you don't write a specific shutdown script, then your program will simply get killed with SIGTERM (and then SIGKILL if it ignores it), which might be unclean. I don't think the case statement I demonstrated is particularly painful to implement.

MK I tried your solution but when I restarted my system ubuntu never starts while the process is the only thing on the screen. How would I get Ubuntu to start normally and still have the process running; would I have to follow what brewbuck listed?