#!/bin/bash
# ========================================================================================
# Top Process of CPU Utilization for Nagios
#
# Written by            : Brian Olson (bolson@nagios.com)
# Release               : 1.0
# Creation date                 : 23 October 2017
# Revision date         : 23 October 2017
# Package               : DTB Nagios Plugin
# Description           : Nagios plugin (script) to check and return top CPU consumer.
#                                                 This script has been designed and written on Linux platform,
#                                                 It will aggregate parent and all child processes by process name
#                                                 and return the process name and CPU percent, as well as a status 
#                                                 of warning or critical if percent is beyond the threshold provided
#                                                 by the -w (warning) and -c (critical arguments.
#                                                 
#
# Usage                 : ./check_top [-w <warn>] [-c <crit]

# ----------------------------------------------------------------------------------------
#
# TODO: 
#               
#                
#
# 
# ========================================================================================

# Nagios return codes
STATE_OK=0
STATE_WARNING=1
STATE_CRITICAL=2
STATE_UNKNOWN=3

# Plugin default level
WARNING_THRESHOLD=${WARNING_THRESHOLD:="30"}
CRITICAL_THRESHOLD=${CRITICAL_THRESHOLD:="50"}


# Plugin variable description
PROGNAME=$(basename $0)
RELEASE="Revision 1.0"
AUTHOR="(c) 2017 Brian Olson (bolson@nagios.com)"

# Functions plugin usage
print_release() {
    echo "$RELEASE $AUTHOR"
}

print_usage() {
        echo ""
        echo "$PROGNAME $RELEASE - Top Process of CPU Utilization for Nagios"
        echo ""
        echo "Usage: check_top -w <warning> -c <critical>"
        echo ""
        echo "          -w  Warning level for CPU percent"
	echo "		-c  critical level for CPU percent"
        echo "          -h  Show this page"
        echo ""
    echo "Usage: $PROGNAME"
    echo "Usage: $PROGNAME --help"
    echo "Usage: $PROGNAME -w <warning> -c <critical>"
    echo ""
}

print_help() {
                print_usage
        echo ""
        print_release $PROGNAME $RELEASE
        echo ""
        echo "This plugin will aggregate parent and all child processes and return name of top consumer "
                echo "-w is for reporting warning levels in percent of CPU utilization"
		echo "-c is for reporting critical levels in percent of CPU utilization"
        echo ""
                exit 0
}

# Make sure the correct number of command line arguments have been supplied
#if [ $# -lt 1 ]; then
#    print_usage
#    exit $STATE_UNKNOWN
#fi

# Parse parameters
while [ $# -gt 0 ]; do
    case "$1" in
        -h | --help)
            print_help
            exit $STATE_OK
            ;;
        -v | --version)
                print_release
                exit $STATE_OK
                ;;
        -w | --warning)
                shift
                WARNING_THRESHOLD=$1
                ;;
        -c | --critical)
               shift
                CRITICAL_THRESHOLD=$1
                ;;
        *)  echo "Unknown argument: $1"
            print_usage
            exit $STATE_UNKNOWN
            ;;
        esac
shift
done
# Top CPU Consumer aggregated by process name

vartmp1="$(ps aux | awk '{array[$11]+=$3} END { for (i in array) {print array[i]"\t" i}}' \
| column -t | sort -r | awk '{ if ($1 > 0) { print } }' | head -1 | awk '{print $2 "\t" $1}')"
vartmp2=$(echo $vartmp1 | awk '{print $2}')
varproc=$(echo $vartmp1 | awk '{print $1}')

        if [ -z "$vartmp2" ]; then 
		vartmp2=0
	fi
	varcpu="${vartmp2%.*}"

	if (( $varcpu >= $WARNING_THRESHOLD )) && (( $varcpu < $CRITICAL_THRESHOLD )); then
            label="TOP CPU PROCESS WARNING:"
            result=$STATE_WARNING
        elif (( $varcpu >= $CRITICAL_THRESHOLD )); then
            label="TOP CPU PROCESS CRITICAL:"
            result=$STATE_CRITICAL
        else
            label="TOP CPU PROCESS OK:"
            result=$STATE_OK
        fi
	if (( $varcpu > 0 )); then
		echo -e "$label Process name: $varproc CPU:  $varcpu%"
	else
		echo -e $label " All running processes less than 1%"
	fi
exit $result