path: root/rgmanager/src/resources/nfsserver.sh.in

#!@BASH_SHELL@

#
# NFS Server Script.  Handles starting/stopping Servand doing
# the strange NFS stuff to get it to fail over properly.
#

LC_ALL=C
LANG=C
PATH=/bin:/sbin:/usr/bin:/usr/sbin

V4RECOVERY="/var/lib/nfs/v4recovery"
PROC_V4RECOVERY="/proc/fs/nfsd/nfsv4recoverydir"

export LC_ALL LANG PATH

. $(dirname $0)/ocf-shellfuncs

# SELinux information
which restorecon &> /dev/null && selinuxenabled &> /dev/null
export SELINUX_ENABLED=$?
if [ $SELINUX_ENABLED ]; then
	export SELINUX_LABEL="$(ls -ldZ /var/lib/nfs/statd | cut -f4 -d' ')"
fi

# strip trailing / off so pattern matching will work consistently.
while [ "${OCF_RESKEY_path#${OCF_RESKEY_path%?}}" = "/" ]
do
	OCF_RESKEY_path="${OCF_RESKEY_path%/}"
done

log_do()
{
	ocf_log debug $*
	$* &> /dev/null
	ret=$?
	if [ $ret -ne 0 ]; then
		ocf_log debug "Failed: $*"
	fi
	return $ret
}


meta_data()
{
	cat <<EOT
<?xml version="1.0" ?>
<!DOCTYPE resource-agent SYSTEM "ra-api-1-modified.dtd">
<resource-agent name="nfsserver" version="rgmanager 2.0">
    <version>1.0</version>

    <longdesc lang="en">
        This defines an NFS server resource.  The NFS server
	resource is useful for exporting NFSv4 file systems
	to clients.  Because of the way NFSv4 works, only
	one NFSv4 resource may exist on a server at a
	time.  Additionally, it is not possible to use
	the nfsserver resource when also using local instances
	of NFS on each cluster node.
    </longdesc>

    <shortdesc lang="en">
        This defines an NFS server resource.
    </shortdesc>

    <parameters>
        <parameter name="name" primary="1">
            <longdesc lang="en">
                Descriptive name for this server.  Generally, only
                one server is ever defined per service.
            </longdesc>
            <shortdesc lang="en">
                Name
            </shortdesc>
	    <content type="string"/>
        </parameter>

        <parameter name="path" inherit="mountpoint">
            <longdesc lang="en">
	        This is the path you intend to export.  Usually, this is
		left blank, and the mountpoint of the parent file system
		is used.  This path is passed to nfsclient resources as
		the export path when exportfs is called.
            </longdesc>
            <shortdesc lang="en">
	    	This is the path you intend to export.
            </shortdesc>
	    <content type="string"/>
        </parameter>

        <parameter name="nfspath">
            <longdesc lang="en">
	        This is the path containing shared NFS information which
		is used for NFS recovery after a failover.  This
		is relative to the export path, and defaults to
		".clumanager/nfs".
            </longdesc>
            <shortdesc lang="en">
	        This is the path containing shared NFS recovery
		information, relative to the path parameter.
            </shortdesc>
	    <content type="string" default=".clumanager/nfs"/>
        </parameter>

        <parameter name="statdport">
            <longdesc lang="en">
		Specifies the port number used for RPC listener sockets. If
		this option is not specified, rpc.statd chooses a random
		ephemeral port for each listener socket. This option can be
		used to fix the port value of its listeners when SM_NOTIFY
		requests must traverse a firewall between
		clients and servers.
            </longdesc>
            <shortdesc lang="en">
                This is the port where rpc.statd should listen on.
            </shortdesc>
	    <content type="integer" default=""/>
        </parameter>

        <parameter name="krbhost">
            <longdesc lang="en">
                This is the Kerberos hostname, which should be set according
                to the floating IP.
            </longdesc>
            <shortdesc lang="en">
                This is the Kerberos hostname.
            </shortdesc>
            <content type="string"/>
        </parameter>

    </parameters>

    <actions>
        <action name="start" timeout="5"/>
	<action name="stop" timeout="5"/>
	<action name="recover" timeout="5"/>

	<action name="status" timeout="5" interval="30"/>
	<action name="monitor" timeout="5" interval="30"/>

	<action name="meta-data" timeout="5"/>
	<action name="validate-all" timeout="30"/>
    </actions>

    <special tag="rgmanager">
    	<attributes maxinstances="1"/>
	<child type="nfsexport" forbid="1"/>
	<child type="nfsserver" forbid="1"/>
	<child type="nfsclient" start="1" stop="2"/>
	<child type="ip" start="2" stop="1"/>
    </special>

</resource-agent>
EOT
}


verify_path()
{
	if [ -z "$OCF_RESKEY_path" ]; then
		ocf_log err "No server path specified."
		return $OCF_ERR_ARGS
	fi

	[ -d "$OCF_RESKEY_path" ] && return 0

	ocf_log err "$OCF_RESKEY_path is not a directory"
	
	return $OCF_ERR_ARGS
}


verify_nfspath()
{
	if [ -z "$OCF_RESKEY_nfspath" ]; then
		echo No NFS data path specified.
		return 1
	fi

	[ -d "$OCF_RESKEY_path" ] && return 0
	
	# xxx do nothing for now.
	return 0
}


verify_statdport()
{
	if [ -z "$OCF_RESKEY_statdport" ]; then
		# this is fine, statdport is optional
		return 0
	fi

	[ $OCF_RESKEY_statdport -gt 0 && $OCF_RESKEY_statdport -le 65535 ] && return 0

	ocf_log err "$OCF_RESKEY_statdport is not a valid port number"

	return $OCF_ERR_ARGS
}


verify_all()
{
	verify_path || return 1
	verify_nfspath || return 1
	verify_statdport || return 1

	return 0
}


nfs_daemons()
{
	declare oper
	declare val

	case $1 in
	start)
		ocf_log info "Starting NFS daemons"
		if [ -z "$OCF_RESKEY_krbhost" ]; then
			/etc/init.d/nfs start
			rv=$?
		else
			unshare -u @BASH_SHELL@ -c "hostname $OCF_RESKEY_krbhost; /etc/init.d/nfs start"
			rv=$?
			unshare -u @BASH_SHELL@ -c "hostname $OCF_RESKEY_krbhost; /etc/init.d/rpcgssd start"
			if [ $rv -ne 0 ]; then
				ocf_log err "Failed to start rpcgssd"
				return $OCF_ERR_GENERIC
			fi
			unshare -u @BASH_SHELL@ -c "hostname $OCF_RESKEY_krbhost; /etc/init.d/rpcidmapd start"
			if [ $rv -ne 0 ]; then
				ocf_log err "Failed to start rpcidmapd"
				return $OCF_ERR_GENERIC
			fi
		fi

		if [ $rv -ne 0 ]; then
			ocf_log err "Failed to start NFS daemons"
			return $OCF_ERR_GENERIC
		fi

		ocf_log debug "NFS daemons are running"
		return $OCF_SUCCESS
		;;
	stop)
		ocf_log info "Stopping NFS daemons"
		if [ -n "$OCF_RESKEY_krbhost"]; then
			if ! /etc/init.d/rpcidmapd stop; then
				ocf_log err "Failed to stop rpcidmapd"
				return $OCF_ERR_GENERIC
			fi
			if ! /etc/init.d/rpcgssd stop; then
				ocf_log err "Failed to stop rpcgssd"
				return $OCF_ERR_GENERIC
			fi
		fi

		if ! /etc/init.d/nfs stop; then
			ocf_log err "Failed to stop NFS daemons"
			return $OCF_ERR_GENERIC
		fi

		ocf_log debug "NFS daemons are stopped"

		return $OCF_SUCCESS
		;;
	status|monitor)
		declare recoverydir="$OCF_RESKEY_path/$OCF_RESKEY_nfspath/v4recovery"
		val=$(cat $PROC_V4RECOVERY)

		[ "$val" = "$recoverydir" ] || ocf_log warning \
			"NFSv4 recovery directory is $val instead of $recoverydir"
		/etc/init.d/nfs status
		if [ $? -ne 0 ]; then
			ocf_log err "NFS is not running"
			return $OCF_NOT_RUNNING
		fi
		/etc/init.d/rpcgssd status
		if [ $? -ne 0 ]; then
			ocf_log err "rpcgssd is not running"
			return $OCF_NOT_RUNNING
		fi
		/etc/init.d/rpcidmapd status
		if [ $? -ne 0 ]; then
			ocf_log err "rpcidmapd is not running"
			return $OCF_NOT_RUNNING
		fi

		ocf_log debug "NFS daemons are running"
		return $OCF_SUCCESS
		;;
	esac
}


create_tree()
{
	declare fp="$OCF_RESKEY_path/$OCF_RESKEY_nfspath"

	[ -d "$fp" ] || mkdir -p "$fp"

	[ -d "$fp/statd" ] || mkdir -p "$fp/statd"
	[ -d "$fp/v4recovery" ] || mkdir -p "$fp/v4recovery"

	#
	# Create our own private copy which we use for notifies.
	# This way, we can be sure to advertise on possibly multiple
	# IP addresses.
	#
	[ -d "$fp/statd/sm" ] || mkdir -p "$fp/statd/sm"
	[ -d "$fp/statd/sm.bak" ] || mkdir -p "$fp/statd/sm.bak"
	[ -d "$fp/statd/sm-ha" ] || mkdir -p "$fp/statd/sm-ha"
	[ -n "`id -u rpcuser`" -a "`id -g rpcuser`" ] && chown -R rpcuser.rpcuser "$fp/statd"

	# Create if they don't exist
	[ -f "$fp/etab" ] || touch "$fp/etab"
	[ -f "$fp/xtab" ] || touch "$fp/xtab"
	[ -f "$fp/rmtab" ] || touch "$fp/rmtab"

	[ $SELINUX_ENABLED ] && chcon -R "$SELINUX_LABEL" "$fp"

        #
        # Generate a random state file.  If this ends up being what a client
        # already has in its list, that's bad, but the chances of this
        # are small - and relocations should be rare.
        #
        dd if=/dev/urandom of=$fp/state bs=1 count=4 &> /dev/null
	[ -n "`id -u rpcuser`" -a "`id -g rpcuser`" ] && chown rpcuser.rpcuser "$fp/state"
}

setup_v4recovery()
{
	declare recoverydir="$OCF_RESKEY_path/$OCF_RESKEY_nfspath/v4recovery"

	# mounts /proc/fs/nfsd for us
	lsmod | grep -q nfsd
	if [ $? -ne 0 ]; then
		modprobe nfsd
	fi

	val=$(cat "$PROC_V4RECOVERY")

	# Ensure start-after-start works
	if [ "$val" = "$recoverydir" ]; then
		return 0
	fi

	#
	# If the value is not default, there may be another
	# cluster service here already which has replaced
	# the v4 recovery directory.  In that case,
	# we must refuse to go any further.
	#
	if [ "$val" != "$V4RECOVERY" ]; then
		ocf_log err "NFSv4 recovery directory has an unexpected value: $val"
		return 1
	fi

	#
	# Redirect nfs v4 recovery dir to shared storage
	#
	echo "$recoverydir" > "$PROC_V4RECOVERY"
	if [ $? -ne 0 ]; then
		echo "Uh oh... echo failed!?"
	fi

	val="$(cat $PROC_V4RECOVERY)"
	if [ "$val" != "$recoverydir" ]; then
		ocf_log err "Failed to change NFSv4 recovery path"
		ocf_log err "Wanted: $recoverydir; got $val"
		return 1
	fi

	return 0
}


cleanup_v4recovery()
{
	#
	# Restore nfsv4 recovery directory to default
	#
	echo "$V4RECOVERY" > "$PROC_V4RECOVERY"
	return $?
}


is_bound()
{
	mount | grep -q "$1 on $2 type none (.*bind.*)"
	return $?
}


setup_tree()
{
	declare fp="$OCF_RESKEY_path/$OCF_RESKEY_nfspath"

	if is_bound $fp/statd /var/lib/nfs/statd; then
		ocf_log debug "$fp is already bound to /var/lib/nfs/statd"
		return 0
	fi

	mount -o bind "$fp/statd" /var/lib/nfs/statd
	cp -a "$fp"/*tab /var/lib/nfs
	[ $SELINUX_ENABLED ] && restorecon /var/lib/nfs
}


cleanup_tree()
{
	declare fp="$OCF_RESKEY_path/$OCF_RESKEY_nfspath"

	if is_bound "$fp/statd" /var/lib/nfs/statd; then
		log_do umount /var/lib/nfs/statd || return 1
	else
		ocf_log debug "$fp is not bound to /var/lib/nfs/statd"
	fi

	cp -a /var/lib/nfs/*tab "$fp"

	return 0
}

start_locking()
{
	declare ret
	declare statdport=""
	[ -x /sbin/rpc.statd ] || return 1
	
	#
	# Synchronize these before starting statd
	#
	cp -f /var/lib/nfs/statd/sm-ha/* /var/lib/nfs/statd/sm 2> /dev/null
	cp -f /var/lib/nfs/statd/sm/* /var/lib/nfs/statd/sm-ha 2> /dev/null

	if pidof rpc.statd &> /dev/null; then
		ocf_log debug "rpc.statd is already running"
		return 0
	fi

	if [ -n "$OCF_RESKEY_statdport" ]; then
		statdport="-p $OCF_RESKEY_statdport"
	fi

	#
	# Set this resrouce script as the callout program.  We are evil.
	# In cases where we want to preserve lock information, this is needed
	# because we can't do the "copy" that we do on the down-state...
	#
	ocf_log info "Starting rpc.statd"
	rm -f /var/run/sm-notify.pid
	rpc.statd -H $0 -d $statdport
	ret=$?
	if [ $ret -ne 0 ]; then
		ocf_log err "Failed to start rpc.statd"
		return $ret
	fi
	touch /var/lock/subsys/nfslock
	return $ret
}


terminate()
{
	declare pids
	declare i=0

	while : ; do
		pids=$(pidof $1)
		[ -z "$pids" ] && return 0
	 	kill $pids
		sleep 1
		((i++))
		[ $i -gt 3 ] && return 1
	done
}


killkill()
{
	declare pids
	declare i=0

	while : ; do
		pids=$(pidof $1)
		[ -z "$pids" ] && return 0
	 	kill -9 $pids
		sleep 1
		((i++))
		[ $i -gt 3 ] && return 1
	done
}

stop_process()
{
	declare process=$1

	ocf_log info "Stopping $process"
	if terminate $process; then
		ocf_log debug "$process is stopped"
	else
		if killkill $process; then
			ocf_log debug "$process is stopped"
		else
			ocf_log debug "Failed to stop $process"
			return 1
		fi
	fi
	return 0
}

stop_locking()
{
	ret=0

	# sm-notify can prevent umount of /var/lib/nfs/statd if
	# it is still trying to notify unresponsive clients.
	stop_process sm-notify
	if [ $? -ne 0 ]; then
		ret=1
	fi

	stop_process rpc.statd
	if [ $? -ne 0 ]; then
		ret=1
	fi

	return $ret
}


case $1 in
start)
	# Check for and source configuration file
	ocf_log info "Starting NFS Server $OCF_RESKEY_name"
	create_tree || exit 1
	setup_tree || exit 1
	setup_v4recovery || exit 1

	start_locking
	nfs_daemons start
	rv=$?
	if [ $rv -eq 0 ]; then
		ocf_log info "Started NFS Server $OCF_RESKEY_name"
		exit 0
	fi

	ocf_log err "Failed to start NFS Server $OCF_RESKEY_name"
	exit $rv
	;;

status|monitor)
	nfs_daemons status 
	exit $?
	;;
		    
stop)
	if ! nfs_daemons stop; then
		ocf_log err "Failed to stop NFS Server $OCF_RESKEY_name"
		exit $OCF_ERR_GENERIC
	fi

	# Copy the current notify list into our private area
	ocf_log debug "Copying sm files for future notification..."
	rm -f /var/lib/nfs/statd/sm-ha/* &> /dev/null
	cp -f /var/lib/nfs/statd/sm/* /var/lib/nfs/statd/sm-ha &> /dev/null

	stop_locking || exit 1
	cleanup_v4recovery
	cleanup_tree || exit 1
	exit 0
	;;

add-client)
	ocf_log debug "$0 $1 $2 $3"
	touch /var/lib/nfs/statd/sm/$2
	touch /var/lib/nfs/statd/sm-ha/$2
	exit 0
	;;

del-client)
	ocf_log debug "$0 $1 $2 $3"
	touch /var/lib/nfs/statd/sm/$2
	rm -f /var/lib/nfs/statd/sm-ha/$2
	exit 0
	;;

recover|restart)
	$0 stop || exit $OCF_ERR_GENERIC
	$0 start || exit $OCF_ERR_GENERIC
	exit 0
	;;

meta-data)
	meta_data
	exit 0
	;;

validate-all)
	verify_all
	exit $?
	;;
*)
	echo "usage: $0 {start|stop|status|monitor|restart|recover|add-client|del-client|meta-data|validate-all}"
	exit $OCF_ERR_UNIMPLEMENTED
	;;
esac

exit 0