#! /bin/sh

# Author: Petter Reinholdtsen <pere@hungry.com>
# License: GNU General Public License v2 or later
#
### BEGIN INIT INFO
# Provides:          megaide-statusd
# Required-Start:    $local_fs $remote_fs
# Required-Stop:     $local_fs $remote_fs
# Default-Start:     2 3 4 5
# Default-Stop:      1 0 6
# Short-Description: Check megaide-status values in the background.
### END INIT INFO

set -e

PATH=/sbin:/bin:/usr/sbin:/usr/bin
DESC="megaide-status monitor"
NAME=megaide-statusd
PIDFILE=/var/run/$NAME.pid
STATUSFILE=/var/run/$NAME.status
SCRIPTNAME=/etc/init.d/$NAME


# Do not touch you can configure this in /etc/default/megaide-statusd
MAILTO=root   # Where to report problems
PERIOD=600    # Seconds between each check    (default 10 minutes)
REMIND=7200   # Seconds between each reminder (default 2 hours)
RUN_DAEMON=yes

[ -e /etc/default/megaide-statusd ] && . /etc/default/megaide-statusd

# Gracefully exit if the package has been removed.
test -x /usr/sbin/megaide-status || exit 0

. /lib/lsb/init-functions
[ -e /etc/default/rcS ] && . /etc/default/rcS

if [ $RUN_DAEMON = "no" ] ; then
	log_begin_msg "megaide-statusd is disabled in /etc/default/megaide-statusd, not starting."
    	log_end_msg 0
	exit 0
fi

check_megaide() {
    echo $$ > $PIDFILE.new && mv $PIDFILE.new $PIDFILE
    while true ; do
        # Check ever $PERIOD seconds, send email on every status
        # change and repeat ever $REMIND seconds if the raid is still
        # bad.
        if (megaide-status) |grep -q 'NOT OPTIMAL' ; then
            BADRAID=true
            logger -t megaide-statusd "detected non-optimal RAID status"
        else
            BADRAID=false
        fi
        STATUSCHANGE=false
        if [ true = "$BADRAID" ] ; then
            # RAID not OK
            (megaide-status) > $STATUSFILE.new
            if [ ! -f $STATUSFILE ] ; then # RAID just became broken
                STATUSCHANGE=true
                mv $STATUSFILE.new $STATUSFILE
            elif cmp -s $STATUSFILE $STATUSFILE.new ; then
                # No change.  Should we send reminder?
                LASTTIME="`stat -c '%Z' $STATUSFILE`"
                NOW="`date +%s`"
                SINCELAST="`expr $NOW - $LASTTIME`"
                if [ $REMIND -le "$SINCELAST" ]; then
                    # Time to send reminder
                    STATUSCHANGE=true
                    mv $STATUSFILE.new $STATUSFILE
                else
                    rm $STATUSFILE.new
                fi
            else
                STATUSCHANGE=true
                mv $STATUSFILE.new $STATUSFILE
            fi
        else
            # RAID OK
            if [ -f $STATUSFILE ] ; then
                rm $STATUSFILE
                STATUSCHANGE=true
            fi
        fi
                
        if [ true = "$STATUSCHANGE" ]; then
            hostname="`uname -n`"
            (
                cat <<EOF 
This is a RAID status update from megaide-statusd.
The megaide-status program reports that one of the RAIDs changed state:

EOF
                if [ -f $STATUSFILE ] ; then
                    cat $STATUSFILE
                else
                    (megaide-status)
                fi
                echo
                echo "Report from $0 on $hostname"
            ) | mail -s "info: MegaRAID raid status change on $hostname" $MAILTO
        fi
        sleep $PERIOD
    done
}

check_daemon() {
        # Let's check if there is a daemon which is really running and not timing out
        DAEMON_RUN=`ps aux | grep "/etc/init.d/megaide-statusd check_megaide" | grep -v grep | grep -v daemon`
        if [ -n "$DAEMON_RUN" ] ; then
                return 1;
        else
                return 0;
        fi
}

#
#       Function that starts the daemon/service.
#
d_start() {
    [ -f $PIDFILE ] && PID="`cat $PIDFILE`"
    if [ "$PID" ] ; then
        log_progress_msg "Daemon already running. Refusing to start another"
        return 0
    elif check_daemon ; then
        # Use the daemon package to turn this script into a daemon
        start-stop-daemon --start --quiet --pidfile $PIDFILE \
                --oknodo --exec /usr/bin/daemon $SCRIPTNAME check_megaide
	return 0
    else
        log_progress_msg "Daemon is already running. Refusing to start another"
        return 0
    fi
}

#
#       Function that stops the daemon/service.
#
d_stop() {
        if [ -f $PIDFILE ] ; then
                # Doesn't work (kill init script instance, but not daemon...)
                #start-stop-daemon --oknodo --stop --quiet --pidfile $PIDFILE 
                DAEMONPID=`ps aux | grep '/usr/bin/daemon /etc/init.d/megaide-statusd check_megaide' | grep -v 'grep' | awk '{ print $2 }'`
                SCRIPTPID=`cat $PIDFILE`
                kill -9 $DAEMONPID $SCRIPTPID || true
                rm -f $PIDFILE
                return 0
        else
                log_progress_msg "Daemon is already stopped."
                return 0
        fi
}

# This is a workaround function which does not directly exit and
# therefore can be used by a restart
d_stop_by_restart() {
        if [ -f $PIDFILE ] ; then
                # Doesn't work (kill init script instance, but not daemon...)
                #start-stop-daemon --oknodo --stop --quiet --pidfile $PIDFILE
                DAEMONPID=`ps aux | grep '/usr/bin/daemon /etc/init.d/megaide-statusd check_megaide' | grep -v 'grep' | awk '{ print $2 }'`
                SCRIPTPID=`cat $PIDFILE`
                kill -9 $DAEMONPID $SCRIPTPID || true
                rm -f $PIDFILE
                log_end_msg 0
        else
                log_progress_msg "Daemon is already stopped."
                log_end_msg 0
        fi
}

case "$1" in
  start)
        echo -n ""
        log_begin_msg "Starting $DESC: $NAME"
        d_start ; CODE=$?
        log_end_msg $CODE
        ;;
  stop)
        log_begin_msg "Stopping $DESC: $NAME"
        d_stop ; CODE=$?
        log_end_msg $CODE
        ;;
  check_megaide)
        check_megaide
        ;;
  restart|force-reload)
        log_begin_msg "Restarting $DESC: $NAME"
        d_stop_by_restart
        sleep 1
        d_start || CODE=$?
        log_end_msg $CODE
        ;;
  *)
        # echo "Usage: $SCRIPTNAME {start|stop|restart|reload|force-reload}" >&2
        echo "Usage: $SCRIPTNAME {start|stop|restart|force-reload}" >&2
        exit 1
        ;;
esac

exit 0