Basically what i want to accomplish, is some sort of script or method for me to start a node.js socket server script, as a service.
This is to make it so that i don't have to physically run 'node server.js' in SSH and have to sit there with it open.
Any help would be appreciated.
Thanks Scott
I don't see anything criminal in wanting to use something different except forever. In my project i also avoid using this kind of tools and more rely on system capabilities. Since i also try to avoid running my application as root, i cannot use SystemV nor Upstart.
And here comes mighty shell scripting! I have created few bash scripts that are doing simply tasks, such as process watchdog with ability to start, stop, restart and query status.
Check this code. Feel free to modify it as you will. To use it -- put your command there in COMMAND parameter. And execute ./path_to_script.sh -start. This will create watchdog process, which will start your node process and watch if it dies or not and if yes, it will restart it. It is far not ideal, so if anyone has something to fix, add, remove here, feel free to comment below.
#!/bin/bash
CURRENT_PATH=$(pwd)
LOGFOLDER=$CURRENT_PATH"/logs/"
PIDFOLDER=$CURRENT_PATH"/pid/"
#PID file where the this script process ID is stored
WATCHDOGPIDFILE=$PIDFOLDER"watchdog-admin.pid"
#PID file where the node process ID is stored
NODEPIDFILE=$PIDFOLDER"node-admin.pid"
#Watchdog process error log file
WATCHDOGLOGFILE=$LOGFOLDER"admin-watchdog-error.log"
#Node process error log file
NODELOGFILE=$LOGFOLDER"admin-error.log"
#Command to be executed on daemon start
COMMAND="node ./admin/app.js 1> /dev/null 2>> $NODELOGFILE"
ARG_1=$1
start() {
if [ -e $NODEPIDFILE ]; then
PID=$(cat $NODEPIDFILE)
if [ $(ps -o pid | grep $PID) ]; then
return;
else
touch $NODEPIDFILE
nohup $COMMAND &
echo $! > $NODEPIDFILE
fi
else
touch $NODEPIDFILE
nohup $COMMAND &
echo $! > $NODEPIDFILE
fi
}
stop() {
if [ -e $NODEPIDFILE ]; then
PID=$(cat $NODEPIDFILE)
if [ $(ps -o pid | grep $PID) ]; then
kill -9 $PID
fi
rm $NODEPIDFILE
fi
}
stopdaemon() {
stop
rm $WATCHDOGPIDFILE
exit 0
}
log() {
echo $1 >> $WATCHDOGLOGFILE
}
keep_alive() {
if [ -e $NODEPIDFILE ]; then
PID=$(cat $NODEPIDFILE)
if [ $(ps -o pid | grep $PID) ]; then
return;
else
log "Jim, he is dead!! Trying ressurection spell..."
start
fi
else
start
fi
}
case x${ARG_1} in
x-start )
echo "Starting daemon watchdog"
nohup "$0" -daemon &> /dev/null &
;;
x-daemon )
if [ -e $WATCHDOGPIDFILE ]; then
PID=$(cat $WATCHDOGPIDFILE)
if [ $(ps -o pid | grep $PID) ]; then
exit 0;
fi
fi
touch $WATCHDOGPIDFILE
echo $$ > $WATCHDOGPIDFILE
#trap the interruption or kill signal
trap stopdaemon INT SIGINT TERM SIGTERM
start
while true; do
keep_alive
wait
sleep 1
done
;;
x-stop )
echo "Stopping daemon watchdog"
PID=$(cat $WATCHDOGPIDFILE)
kill $PID
;;
x-status )
#check if process is running and PID file exists, and report it back
;;
x )
echo "Usage {start|stop|status}"
esac
exit 0
It sounds like you want a tool like forever. There's a getting started blog post here.
There are also more generic tools like monit and upstart.
I am personally using forever on my side project and it has worked just fine.
This gets complicated quickly. What you are really asking for is how to "daemonize" a process. It's more than simply running it in the background. Ideally you want to be running it in such a way that it doesn't tie down removable filesystems, and you want to be keeping its log output and not having it spitting to your old terminal. You don't want it accidentally killed when your shell exits, etc...
If you are on a Debian-based system like Ubuntu, the /sbin/start-stop-daemon does much of what you want. Writing an Ubuntu Upstart configuration for it was mentioned, and that's an option too. On other distros like Fedora, systemd provides what is generally considered the "gold plated" version of this automatic daemonization service. It can daemonize and manage basically any program that isn't explicitly interactive.
