Application launchers

Application launchers are installed in the directory pointed by /opt/opensvc/etc/svcnamed.d. They follow the SysV guidelines:

Application can be started and stopped without touching the other service resources with:

/opt/opensvc/etc/myservice stopapp
/opt/opensvc/etc/myservice startapp

The OPENSVC_SVCNAME variable is set by OpenSVC to the service name, so that developers can write reusable launcher scripts. Examples of this variable usage below.

Common launchers

Crontab handler

Services often come with specific schedules. Either they start their own private scheduler, or they use the main system scheduler. In the later case, services' cron jobs must be installed upon service start and removed upon service stop to make sure the jobs don't get scheduled on multiple nodes. The following launcher script can be used to do that on hosts with split crontab support. It installs and removes the crontabs in /opt/opensvc/etc/myservice.d/cron.d/.

#!/bin/bash

cron_d=/$OPENSVC_SVCNAME/etc/init.d/cron.d
cf=

[ "$OPENSVC_SVCNAME" = "" ] && {
	echo "\$OPENSVC_SVCNAME must be set"
	exit 1
}

function check_crontab {
        find $cron_d/$1 -user root -a -group root -a -perm -o-w | grep $1 >/dev/null 2>&1 && return 0
	cat $cron_d/$1 | cut -f6 | grep root >/dev/null 2>&1 && return 1
        return 0
}

function add_crontab {
	if [ "${cf/$1//}" != "$cf" ] ; then
		echo $cf
                return
	fi
	if ! check_crontab $1 ; then
		echo "discard $1 because of security risks" >&2
		echo $cf
                return
	fi
	echo "$cf $1"
}

function get_crontabs {
	for c in $(ls $cron_d | grep -v '\.') ; do
		if [ -f $c.$HOSTNAME ] ; then
			add_crontab $c.$HOSTNAME
		else
			add_crontab $c
		fi
	done
}

case $1 in
start)
        for cf in $(get_crontabs) ; do
		cp -f $cron_d/$cf /etc/cron.d/$OPENSVC_SVCNAME.$cf
                chown root:root /etc/cron.d/$OPENSVC_SVCNAME.$cf
                echo "installed $cf as /etc/cron.d/$OPENSVC_SVCNAME.$cf"
        done
	;;
stop)
	rm -f /etc/cron.d/$OPENSVC_SVCNAME.*
	;;
esac

Note that an additional safety net is needed to keep the cron jobs from failing after a node unclean reboot: all cron jobs must verify the presence of their executable file. Example:

22 23 * * * [ -x /myservice/app/jobs/job1 ] && /myservice/app/jobs/job1
23 23 * * * [ -x /myservice/app/jobs/job2 ] && /myservice/app/jobs/job2