#!/bin/sh
#
# fence_sanlockd - daemon for fence_sanlock agent
#
# chkconfig: 2345 20 80
# description: starts and stops fence_sanlockd
#

### BEGIN INIT INFO
# Provides: fence_sanlockd
# Required-Start: $time $syslog $remote_fs
# Required-Stop: $syslog
# Should-Start:
# Should-Stop:
# Default-Start: 2 3 4 5
# Default-Stop: 0 1 6
# Short-Description: starts and stops fence_sanlockd
# Description: starts and stops fence_sanlockd
### END INIT INFO

prog="fence_sanlockd"
agent="fence_sanlock"
daemonrundir="/var/run/$prog"
agentrundir="/var/run/$agent"
runfile="$daemonrundir/$prog.pid"
fifofile="$daemonrundir/$prog.fifo"
lockfile="/var/lock/subsys/$prog"
exec="/usr/sbin/$prog"

FENCESANLOCKDOPTS="-w"

test -x $exec || { echo "$exec not installed";
	 if [ "$1" = "stop" ]; then exit 0;
	 else exit 5; fi; }

[ -f /etc/sysconfig/$prog ] && . /etc/sysconfig/$prog

. /etc/rc.status
rc_reset

case "$1" in
start)
	# start wdmd and sanlock daemons if they aren't running
	service wdmd status > /dev/null 2>&1 || service wdmd start
	service sanlock status > /dev/null 2>&1 || service sanlock start

	if [ ! -d $daemonrundir ]; then
		install -d -m 775 $daemonrundir
		[ -x /sbin/restorecon ] && restorecon $daemonrundir
	fi
	if [ ! -d $agentrundir ]; then
		install -d -m 775 $agentrundir
		[ -x /sbin/restorecon ] && restorecon $agentrundir
	fi

	if [ -e $runfile ]; then
		if checkproc $exec ; then
			echo -n "$prog is already running."
			rc_status -v
			rc_exit
		else
			echo "Removing stale PID file $runfile."
			rm -f $runfile
		fi
	fi

	echo -n "Starting $prog "
	startproc $exec daemon $FENCESANLOCKDOPTS
	rc_status -v
	;;
stop)
	agent_ps="$(ps ax -o pid,args | grep $agent | grep -v grep | grep -v $prog)"

	[ -n "$agent_ps" ] && {
		agent_pid="$(echo $agent_ps | awk '{print $1}')"
		echo -n "cannot stop while $agent $agent_pid is running"
		rc_failed 1
		rc_status -v
		rc_exit
	}

	# Ideally, we'd like a general way to check if anything
	# needs fencing to continue running, but without that,
	# check what we know, which is that dlm requires it.
	if [ -d /sys/kernel/dlm/ ]; then
		count="$(ls -A /sys/kernel/dlm/ | wc -l)"
		if [ $count -ne 0 ]; then
			echo -n "cannot stop while dlm lockspaces exist"
			rc_failed 1
			rc_status -v
			rc_exit
		fi
	fi

	if [ -d /sys/kernel/config/dlm/cluster ]; then
		# this dir exists while dlm_controld is running
		echo -n "cannot stop while dlm is running"
		rc_failed 1
		rc_status -v
		rc_exit
	fi

	PID=$(pidofproc -p $runfile $prog)

	# We have to use SIGHUP to mean stop because sanlock
	# uses SIGTERM to mean that the lockspace failed.
	echo -n $"Sending stop signal $prog ($PID)"
	killproc -p $runfile $prog -HUP
	retval=$?
	if [ $retval -ne 0 ]; then
		rc_failed 1
		rc_status -v
		rc_exit
	fi

	# fence_sanlockd won't see the SIGHUP if it's
	# still waiting for config from the fifo, so
	# send invalid config to the fifo to make it fail.
	if [ -p $fifofile ]; then
		echo "" > $fifofile
	fi

	echo -n $"Waiting for $prog ($PID) to stop"
	timeout=10
	while checkpid $PID; do
		sleep 1
		timeout=$((timeout - 1))
		if [ "$timeout" -le 0 ]; then
			echo -n "failed waiting for $prog ($PID) to stop"
			rc_failed 1
			rc_status -v
			rc_exit
		fi
	done
	
	rm -f $lockfile

	# stop wdmd and sanlock daemons if they are running
	service sanlock status > /dev/null 2>&1 && service sanlock stop
	service wdmd status > /dev/null 2>&1 && service wdmd stop

	rc_status -v
	;;
try-restart)
	$0 status >/dev/null &&  $0 restart
	rc_status
	;;
restart)
	$0 stop
	$0 start
	rc_status
	;;
reload)
	killproc -HUP $exec
	rc_status -v
	;;
status)
	echo -n "Checking status of $prog "
	checkproc $exec
	rc_status -v
	;;
*)
	echo "Usage: $0 {start|stop|restart|try-restart|reload|status}"
	rc_failed 2
	rc_exit
	;;
esac
rc_exit