#!/bin/bash
#
# Resource script for icinga/nagios daemon
#
# Description:  Manages icinga/nagios daemon as an OCF resource in 
#               an High Availability setup.
#
# Author:       Lennart Betz <lennart.betz@netways.de>
# License:      GNU General Public License (GPL) 
# Copyright:    (C) NETWAYS GmbH
#
# Changelog:
#
# 2011-08-23	0.1	initial release
# 2011-09-12	0.2	change parameter config to configfile
#			add parameter envfiles
# 2012-12-20	0.3	add validation of lsof, pgrep, ... exist
#
#	Include /etc/default|sysconfig/icinga|nagios for defaults.
#
#	At this time only tested with debian. Check it on openSuSE, RedHat...
#
#
#
#	usage: $0 {start|stop|status|monitor|validate-all|meta-data}
#
#	The "start" arg starts icinga/nagios.
#
#	The "stop" arg stops it.
#
# OCF parameters:
# OCF_RESKEY_binary
# OCF_RESKEY_configfile
# OCF_RESKEY_pidfile
# OCF_RESKEY_commandfile
# OCF_RESKEY_envfiles
#
##########################################################################
# Initialization:

. ${OCF_ROOT}/resource.d/heartbeat/.ocf-shellfuncs

##########################################################################

OCF_RESKEY_binary_default=/usr/local/icinga/bin/icinga
OCF_RESKEY_configfile_default=/usr/local/icinga/etc/icinga.cfg
OCF_RESKEY_pidfile_default=/usr/local/icinga/var/icinga.lock
OCF_RESKEY_commandfile_default=/usr/local/icinga/var/rw/icinga.cmd

# what commands are used and have to exist
DEPENDENCIES="grep lsof pgrep"

usage() {
  cat <<-!
usage: $0 {start|stop|status|monitor|validate-all|meta-data}

action:
        start   start the ${DAEMON} daemon

        stop    stop the ${DAEMON} daemon

        status  return the status of the ${DAEMON} daemon, running or down

        methods return the set of commands we support

        monitor return TRUE if the ${DAEMON} daemon appears to be working.

        meta-data       show meta data message

        validate-all    validate the instance parameters
	!
  exit $OCF_ERR_ARGS
}


meta_data() 
{
cat <<END
<?xml version="1.0"?>
<!DOCTYPE resource-agent SYSTEM "ra-api-1.dtd">
<resource-agent name="${DAEMON}">
<version>1.0</version>
<longdesc lang="en">
This script manages the ${DAEMON} daemon
</longdesc>
<shortdesc lang="en">OCF Resource Agent compliant ${DAEMON} daemon script.</shortdesc>

<parameters>

<parameter name="binary" unique="0" required="0">
<longdesc lang="en">
The ${DAEMON} binary path.
</longdesc>
<shortdesc lang="en">Full path to the ${DAEMON} binary.</shortdesc>
<content type="string" default="${OCF_RESKEY_binary_default}" />
</parameter>

<parameter name="configfile" unique="0" required="0">
<longdesc lang="en">
The ${DAEMON} daemon configuration file name with full path. 
For example, "${OCF_RESKEY_configfile_default}"
</longdesc>
<shortdesc lang="en">Configuration file name with full path</shortdesc>
<content type="string" default="${OCF_RESKEY_configfile_default}" />
</parameter>

<parameter name="pidfile" unique="0" required="0">
<longdesc lang="en">
The ${DAEMON} daemon pidfile name with full path. Isn't set, use "lock_file" from config file. 
For example, "${OCF_RESKEY_pidfile_default}"
</longdesc>
<shortdesc lang="en">Pidfile name with full path.</shortdesc>
<content type="string" default="${OCF_RESKEY_pidfile_default}" />
</parameter>

<parameter name="comandfile" unique="0" required="0">
<longdesc lang="en">
The ${DAEMON} commandfile with full path. Isn't set, use "command_file" from config file. 
For example, "${OCF_RESKEY_commandfile_default}"
</longdesc>
<shortdesc lang="en">Commandfile with full path.</shortdesc>
<content type="string" default="${OCF_RESKEY_commandfile_default}" />
</parameter>

<parameter name="envfiles">
<longdesc lang="en">
Files (one or more) which contain extra environment variables.
</longdesc>
<shortdesc lang="en">environment settings files</shortdesc>
<content type="string" default="/etc/sysconfig|default/${DAEMON}" />
</parameter>

</parameters>

<actions>
<action name="start" timeout="20"/>
<action name="stop" timeout="20"/>
<action name="monitor" timeout="20s" interval="30s" start-delay="3s" />
<action name="validate-all" timeout="20s"/>
<action name="meta-data"  timeout="5s"/>
</actions>
</resource-agent>
END
exit $OCF_SUCCESS
}

get_pids()
{
	PIDS=( )

	# Seek by pidfile
	PIDS[1]=$(awk '1{print $1}' $PIDFILE 2>/dev/null)
	if [ -n "${PIDS[1]}" ]; then
		typeset exe
		exe=$(ls -l "/proc/${PIDS[1]}/exe" 2>/dev/null)
		if [ $? = 0 ]; then
			exe=${exe##*-> }
			if ! [ "$exe" = ${OCF_RESKEY_binary} ]; then
				PIDS[1]=""
			fi
		else
			PIDS[1]=""
		fi
	fi
	# Seek by command file
	if [ $COMMANDFILE ]; then
		typeset pid
		for p in $(lsof -FpR $COMMANDFILE 2>/dev/null); do
			if [ "X$p" = "XR1" ]; then
				if [ -z ${PIDS[2]} ]; then
					PIDS[2]=$pid
				else
					PIDS[2]="$pid ${PIDS[2]}"
				fi
			fi
			pid=$(echo $p| sed 's/^p//g')
		done
	fi
	# Seek by pattern, parent 1 (init); for loop is need for forks 
	for i in $(pgrep -P 1 -f "$PROCESS_PATTERN"); do
		if [ "X$i" == "X${PIDS[1]}" ]; then
			PIDS[0]=$i
		fi
	done

}

pids_sane()
{
	if [ "${PIDS[1]}" = "${PIDS[2]}" ] && [ "${PIDS[1]}" = "${PIDS[0]}" ]; then
		return $OCF_SUCCESS
	else
		ocf_log err "$DAEMON: PID mismatch"
		return $OCF_ERR_GENERIC
	fi
}

is_dead()
{
	if
		[ -z "${PIDS[0]}" ] &
		[ -z "${PIDS[2]}" ]
	then
		return 0
	else
		return 1
	fi
}

monitor()
{
	get_pids
	#echo ps=${PIDS[0]} pidfile=${PIDS[1]} socket=${PIDS[2]}
	if ! pids_sane; then
		return $OCF_ERR_GENERIC
	fi

	if is_dead; then
		return $OCF_NOT_RUNNING
	fi
	
	return $OCF_SUCCESS
}

status()
{
	monitor
}

start()
{
	status
	retVal=$?
	if [ $retVal -eq $OCF_SUCCESS ]; then
		ocf_log info "${DAEMON} already running"
		exit $OCF_SUCCESS
	elif [ $retVal -ne $OCF_NOT_RUNNING ]; then
		ocf_log err "Error. Unknown status."
		exit $OCF_ERR_GENERIC
	fi

        $PROCESS_PATTERN
        if [ $? -ne 0 ]; then
		ocf_log err "Error. ${DAEMON} returned error $?."
                exit $OCF_ERR_GENERIC
        fi

        ocf_log info "Started ${DAEMON} daemon."
        exit $OCF_SUCCESS
}


stop()
{
	pid=`cat $PIDFILE 2>/dev/null`
	if [ -n "$pid" ] ; then
		kill $pid
		if [ $? -ne 0 ]; then
			kill -SIGKILL $pid
			if [ $? -ne 0 ]; then
				ocf_log err "Error. Could not stop ${DAEMON}."
				return $OCF_ERR_GENERIC
			fi
		fi
		rm $PIDFILE 2>/dev/null
	fi
	kill -SIGKILL $(pgrep -f "$PROCESS_PATTERN") 2>/dev/null

	ocf_log info "Stopped ${DAEMON}."
	exit $OCF_SUCCESS
}

validate_all()
{
	for file in $DEPENDENCIES; do
                if [ ! -x "$(which $file)" ]; then
                        ocf_log err "$DAEMON $file could not be found or isn't executable."
                        exit $OCF_ERR_INSTALLED
                fi
        done

	if [ ! -x $OCF_RESKEY_binary ]; then
		ocf_log err "$DAEMON binary $OCF_RESKEY_binary does not exist or is not executable."
		exit $OCF_ERR_INSTALLED
	fi

	if [ ! -f $OCF_RESKEY_configfile ]; then
		ocf_log err "Config $OCF_RESKEY_configfile does not exist."
		exit $OCF_ERR_CONFIGURED
	fi

	return $OCF_SUCCESS
}

#
# Main
#

: ${OCF_RESKEY_binary=${OCF_RESKEY_binary_default}}
DAEMON=$(basename ${OCF_RESKEY_binary})

if [ -z ${OCF_RESKEY_envfiles} ]; then
	if [ -f /etc/sysconfig/$DAEMON ]; then
		. /etc/sysconfig/$DAEMON
	fi

	if [ -f /etc/default/$DAEMON ]; then
		. /etc/default/$DAEMON
	fi
else
	for i in ${OCF_RESKEY_envfiles}; do
		if [ -f $i ]; then
			. $i
		else
			ocf_log err "Error. Could not read $i, no such file or directory."
			exit $OCF_ERR_CONFIGURED
		fi
	done
fi

: ${OCF_RESKEY_configfile=${OCF_RESKEY_configfile_default}}
: ${OCF_RESKEY_pidfile=${OCF_RESKEY_pidfile_default}}
: ${OCF_RESKEY_socket=${OCF_RESKEY_commandfile_default}}

# if config file is accessable, read pidfile and command file from config file
if [ ! -z ${OCF_RESKEY_configfile} ] && [ -f ${OCF_RESKEY_configfile} ]; then
	# read pidfile from config file
	PIDFILE=$(grep -v "^#" ${OCF_RESKEY_configfile} |grep "^lock_file=" |cut -d "=" -f 2)
	if [ "X$PIDFILE" = "X" ]; then
		ocf_log err "Error. \"lock_file\" entry without whitespaces required in the $DAEMON config file by $DAEMON OCF RA."
		exit $OCF_ERR_GENERIC
	fi
	# read command file from config file
	COMMANDFILE=$(grep -v "^#" ${OCF_RESKEY_configfile} |grep "^command_file=" |cut -d "=" -f 2)
	if [ "X$COMMANDFILE" = "X" ]; then
		ocf_log err "Error. \"command_file\" entry without whitespaces required in the $DAEMON config file by $DAEMON OCF RA."
		exit $OCF_ERR_GENERIC
	fi
else
	PIDFILE=${OCF_RESKEY_pidfile}
	COMMANDFILE=${OCF_RESKEY_commandfile}
fi

PROCESS_PATTERN="${OCF_RESKEY_binary} -d ${OCF_RESKEY_configfile}"

case $1 in
	start)	
		validate_all
		start
		;;
	
	stop)	
		validate_all
		stop
		;;

	status)	
		status
		;;

	monitor)
		monitor
		;;

	validate-all)	
		validate_all
		;;

	meta-data)
		meta_data
		;;

	notify|promote|demote)
		exit $OCF_ERR_UNIMPLEMENTED
		;;
	usage)	usage
		exit $OCF_SUCCESS
		;;

	*)	usage
		exit $OCF_ERR_UNIMPLEMENTED
		;;
esac