I need to install a program as a service in Red Hat. It doesn't background itself, manage its PID file, or manage its own logs. It just runs and prints to STDOUT and STDERR.
Using the standard init scripts as guides, I've developed the following:
#!/bin/bash
#
# /etc/rc.d/init.d/someprog
#
# Starts the someprog daemon
#
# chkconfig: 345 80 20
# description: the someprog daemon
# processname: someprog
# config: /etc/someprog.conf
# Source function library.
. /etc/rc.d/init.d/functions
prog="someprog"
exec="/usr/local/bin/$prog"
[ -e "/etc/sysconfig/$prog" ] && . "/etc/sysconfig/$prog"
lockfile="/var/lock/subsys/$prog"
RETVAL=0
check() {
[ `id -u` = 0 ] || exit 4
test -x "$exec" || exit 5
}
start() {
check
if [ ! -f "$lockfile" ]; then
echo -n $"Starting $prog: "
daemon --user someproguser "$exec"
RETVAL=$?
[ $RETVAL -eq 0 ] && touch "$lockfile"
echo
fi
return $RETVAL
}
stop() {
check
echo -n $"Stopping $prog: "
killproc "exec"
RETVAL=$?
[ $RETVAL -eq 0 ] && rm -f "$lockfile"
echo
return $RETVAL
}
restart() {
stop
start
}
case "$1" in
start)
start
;;
stop)
stop
;;
restart)
restart
;;
status)
status "$prog"
RETVAL=$?
;;
*)
echo $"Usage: $0 {start|stop|restart|status}"
RETVAL=2
esac
exit $RETVAL
It may be that my mistake was to copy-paste and modify some of the existing scripts in /etc/init.d. In any case, the resulting service behaves strangely:
- when I start it with
service someprog start
the program prints to the terminal and the command doesn't complete. - if I CTRL-C, it prints "Session terminated, killing shell… …killed. FAILED". I have to do this to get my shell prompt back again.
- now when I run
service someprog status
it says it's running and lists its PID. I can see it inps
so it is running. - now when I run
service someprog stop
it fails to stop. I can verify that it's still running withps
.
What do I need to change so that someprog
is sent to the background and managed as a service?
Edit: I have now found a couple of related questions, neither of them with an actual answer other than "do something else":
- Call to daemon in a /etc/init.d script is blocking, not running in background
- Getting shell script to run as a daemon on CentOS?
Edit: this answer on double-forking might have solved my problem, but now my program itself double-forks and that works: https://stackoverflow.com/a/9646251/898699
Best Answer
The command "doesn't complete" because the
daemon
function does not run your application in the background for you. You will need to add an&
to the end of yourdaemon
command like so:daemon --user someproguser $exec &
If
someprog
doesn't handleSIGHUP
, you should run the command withnohup
to ensure that your process won't receiveSIGHUP
which tells your process to exit when the parent shell exits. That would look like this:daemon --user someproguser "nohup $exec" &
In your
stop
function,killproc "exec"
isn't doing anything to stop your program. It should read like so:killproc $exec
killproc
requires the full path to your application to stop it properly. I've had some trouble withkillproc
in the past, so you can also just kill the PID in the PIDFILE you should be writingsomeprog
's PID to with something like this:cat $pidfile | xargs kill
You can write the PIDFILE like this:
ps aux | grep $exec | grep -v grep | tr -s " " | cut -d " " -f2 > $pidfile
where
$pidfile
points to/var/run/someprog.pid
.If you want [OK] or [FAILED] on your
stop
function, you should use thesuccess
andfailure
functions from/etc/rc.d/init.d/functions
. You don't need these in thestart
function becausedaemon
calls the appropriate one for you.You also only need quotes around strings with spaces. It's a style choice, though, so it's up to you.
All these changes look like this: