#!/bin/sh
#
# lustre   This shell script takes care of starting and stopping Lustre
#
# chkconfig: - 99 1
# description: Lustre Lite network File System.
#              This starts both Lustre client and server functions.
# processname: lconf
# config: /etc/lustre/config.xml
# pidfile: /var/run/lustre.pid
### BEGIN INIT INFO
# Provides: lustre
# Required-Start: $network +sshd
# Required-Stop: $network
# Should-Start:
# Should-Stop:
# Default-Start: 
# Default-Stop: 0 1 2 3 4 5 6
# Short-Description: Lustre Lite network File System.
# Description: This starts both Lustre client and server functions.
### END INIT INFO


SERVICE=${0##*/}

: ${LUSTRE_CFG:=/etc/lustre/lustre.cfg}
[ -f ${LUSTRE_CFG} ] && . ${LUSTRE_CFG}
[ -f /etc/sysconfig/lustre ] && . /etc/sysconfig/lustre

: ${LUSTRE_CONFIG_XML:=/etc/lustre/config.xml}
: ${LCONF:=lconf}
: ${LCTL:=lctl}
# Some distros use modprobe.conf.local
if [ -f /etc/modprobe.conf.local ]; then
   : ${MODPROBE_CONF:=/etc/modprobe.conf.local}
else
   : ${MODPROBE_CONF:=/etc/modprobe.conf}
fi
# Be sure the proper directories are in PATH. 
export PATH="/sbin:$PATH"

case "$SERVICE" in
    [SK][[:digit:]][[:digit:]]lustre | lustre)
        SERVICE="lustre"
	: ${LCONF_START_ARGS:="${LUSTRE_CONFIG_XML}"}
	: ${LCONF_STOP_ARGS:="--force --cleanup ${LUSTRE_CONFIG_XML}"}
	;;
    *)
	: ${LCONF_START_ARGS:="--group ${SERVICE} --select ${SERVICE}=${HOSTNAME} ${LUSTRE_CONFIG_XML}"}
	: ${LCONF_STOP_ARGS:="--group ${SERVICE} --select ${SERVICE}=${HOSTNAME} --failover --cleanup ${LUSTRE_CONFIG_XML}"}
	;;
esac
LOCK=/var/lock/subsys/$SERVICE

# Source function library.
if [ -f /etc/init.d/functions ] ; then
	. /etc/init.d/functions
fi

# Source networking configuration.
if [ -f /etc/sysconfig/network ] ; then
	. /etc/sysconfig/network
fi

check_start_stop() {
	# Exit codes now LSB compliant
	# Check that networking is up. - exit 'not running'
	[ "${NETWORKING}" = "no" ] && exit 7 

	# exit 'not installed' 
	[ -x ${LCONF} -a -x ${LCTL} ] || exit 5

	if [ ${LUSTRE_CONFIG_XML:0:1} = "/" ] ; then
			if [ ! -f ${LUSTRE_CONFIG_XML} ] ; then
			echo "${0##*/}: Configuration file ${LUSTRE_CONFIG_XML} not found; skipping."
			# exit 'not configured'
			exit 6
		fi
	fi

	# Create /var/lustre directory 
	# This is used by snmp agent for checking lustre services
	#    status online/offline/online pending/offline pending.

	[ -d ${STATUS_DIR:=/var/lustre} ] || mkdir -p $STATUS_DIR
	STATUS=${STATUS_DIR}/sysStatus
}

start() {
	if [ -x "/usr/sbin/clustat" -a "${SERVICE}" = "lustre" ] ; then
		if [ ! -f "/etc/lustre/start-despite-clumanager" ] ; then
		cat >&2 <<EOF
This script was run directly, which can be dangerous if you are using
clumanager to manage Lustre services.

If you are not using clumanager for Lustre services, run the following
command to have this script start Lustre instead:

touch /etc/lustre/start-despite-clumanager
EOF
		RETVAL=6  # program not configured
		return
	    fi
	fi
	check_start_stop
	echo -n "Starting $SERVICE: "
	if [ $UID -ne 0 ]; then
		echo "Lustre should be started as root"
		RETVAL=4 # insufficent privileges
		return
	fi
	# Cat the modprobe file and place all lines that follow a trailing backslash on the same line
	ROUTER=`cat ${MODPROBE_CONF} | sed ':a;N;$!ba;s#\\\[:space:]*\\n##g' | grep lnet | grep forwarding=\"enabled\"`
	if [[ ! -z ${ROUTER} ]]; then
		modprobe lnet
		${LCTL} network configure
	else
		${LCONF} ${LCONF_START_ARGS}
	fi
	RETVAL=$?
	echo $SERVICE
	if [ $RETVAL -eq 0 ]; then
		touch $LOCK
		echo "online" >$STATUS
	else
		echo "online pending" >$STATUS
	fi
}

stop() {
	check_start_stop
	echo -n "Shutting down $SERVICE: "
	if [ $UID -ne 0 ]; then
		echo "Lustre should be stopped as root"
		RETVAL=4 # insufficent privileges
		return
	fi
	# Cat the modprobe file and place all lines that follow a trailing backslash on the same line
+	ROUTER=`cat ${MODPROBE_CONF} | sed ':a;N;$!ba;s#\\\[:space:]*\\n##g' | grep lnet | grep forwarding=\"enabled\"`
	if [[ ! -z ${ROUTER} ]]; then
		MODULE_LOADED=`lsmod | awk ' { print $1 } ' | grep lnet`
		if [[ ! -z ${MODULE_LOADED} ]]; then
			${LCTL} network unconfigure
		fi
		${LCTL} modules | awk '{ print $2 }' | xargs rmmod >/dev/null 2>&1
		# do it again, in case we tried to unload ksocklnd too early
		${LCTL} modules | awk '{ print $2 }' | xargs rmmod

	else
		${LCONF} ${LCONF_STOP_ARGS}
	fi
	RETVAL=$?
	echo $SERVICE
	rm -f $LOCK 
	if [ $RETVAL -eq 0 ]; then
		echo "offline" >$STATUS
	else
		echo "offline pending" >$STATUS
	fi
}

restart() {
	stop
	start
}

status() {
	STATE="stopped"
	# LSB compliance - return 3 if service is not running
	# Lustre-specific returns
	# 150 - partial startup
	# 151 - health_check unhealthy
	# 152 - LBUG
	RETVAL=3
	egrep -q "libcfs|lvfs|portals" /proc/modules && STATE="loaded"

	# check for any routes - on a portals router this is the only thing
	[ "`cat /proc/sys/lnet/routes 2> /dev/null`" ] && STATE="running" && RETVAL=0
	
	# check for any configured devices (may indicate partial startup)
	if [ -d /proc/fs/lustre ]; then
		[ "`cat /proc/fs/lustre/devices 2> /dev/null`" ] && STATE="partial" && RETVAL=150

		# check for either a server or a client filesystem
		MDS="`ls /proc/fs/lustre/mds/*/recovery_status 2> /dev/null`"
		OST="`ls /proc/fs/lustre/obdfilter/*/recovery_status 2> /dev/null`"
		LLITE="`ls /proc/fs/lustre/llite/fs* 2> /dev/null`"
		[ "$MDS" -o "$OST" -o "$LLITE" ] && STATE="running" && RETVAL=0
	else
		# check if this is a router
		if [ -d /proc/sys/lnet ]; then
			ROUTER="`cat /proc/sys/lnet/routes | head -1 | grep -i -c \"Routing enabled\"`"
			if [[ ! -z ${ROUTER} && ${ROUTER} -ge 1 ]]; then
				STATE="running"
				RETVAL=0
			fi
		fi
	fi

	# check for server disconnections 
	DISCON="`grep -v FULL /proc/fs/lustre/*c/*/*server_uuid 2> /dev/null`"
	[ "$DISCON" ] && STATE="disconnected" && RETVAL=0

	# check for servers in recovery
	[ "$MDS$OST" ] && grep -q RECOV $MDS $OST && STATE="recovery" && RETVAL=0

	# check for error in health_check
	HEALTH="/proc/fs/lustre/health_check"
	[ -f "$HEALTH" ] && grep -q "NOT HEALTHY" $HEALTH && STATE="unhealthy" && RETVAL=151

	# check for LBUG
	[ -f  "$HEALTH" ] && grep -q "LBUG" $HEALTH && STATE="LBUG" && RETVAL=152

	# If Lustre is up , check if the service really exists
        # Skip this is we are not checking a specific service
	if [ $RETVAL -eq 0 ] && [ $SERVICE != 'lustre' ]; then
		DUMMY=$( $LCTL dl | grep "$SERVICE")
		[ $? -ne 0 ] && STATE="not_found" && RETVAL=3
	fi

	echo $STATE
}

# See how we were called.
case "$1" in
  start)
	start
	;;
  stop)
	stop
	;;
  restart)
	restart
	;;
  status)
	status $SERVICE
	;;
  *)
	echo "Usage: $SERVICE {start|stop|restart|status}"
	exit 1
esac

exit $RETVAL