#!/bin/sh

# Little Nagios/Icinga/Naemon probe for Linux disks' I/O via /proc/diskstats
#
# Based on https://www.kernel.org/doc/html/latest/admin-guide/iostats.html
#
# Published under Unlicense <http://unlicense.org/> (similar to public domain)

PROGPATH=$( echo $0 | sed -e 's,[\\/][^\\/][^\\/]*$,,' )
REVISION="0.1"

# Stop at the first non-catched error
set -e

# Include check_range()
# Not needed at the moment
#. $PROGPATH/utils.sh
STATE_OK=0
STATE_WARNING=1
STATE_CRITICAL=2
STATE_UNKNOWN=3
STATE_DEPENDENT=4

# Default values
WARNING_STATS=0
CRITICAL_STATS=0

# Output
OUTPUT_EXIT_STATUS=$STATE_OK
OUTPUT_DETAIL_OK=""
OUTPUT_DETAIL_WARNING=""
OUTPUT_DETAIL_CRITICAL=""
OUTPUT_PERFDATA=""

#
# Help function
#
usage() {
	cat <<EOF
Usage :
  $0 [ -i <nb_in_progress_warn_threshold> ] [ -I <nb_in_progress_crit_threshold> ] -d <device_name> [ ... -d ... ]

This probe send statistics about disks on a Linux system.
At the moment, the "number of IO in progress" (field #9) is the only metric
able to generate an alert.

Options :
	-d xx	Device name (usually 'sda'). Can use shell pattern like 'sda*' or 'sd?'.
	-i N	Number of io in progress : warning threshold
	-I N	Number of io in progress : critical threshold

Example :
	This example is a good generic configuration for simple and light systems,
	whether physical or virtual, taking any sda, sdb, vda, vdb, etc. disks found
	and triggering a warning at 1 IO in progress and an alert at 5.
		$0 -w 0 -W 5 -d 'sd?' -d 'vd?'
EOF
}

# Some early checks
if [ ! -r "/proc/diskstats" ]; then
	echo "UNKNOWN file '/proc/diskstats' not found or not readable."
	exit $STATE_UNKNOWN
fi
# TODO: check Linux kernel version. See differences between 2.4+, 2.6+, 4.18+
#       https://www.kernel.org/doc/html/latest/admin-guide/iostats.html

#
# Loop on parameters + tests
#
while getopts hd:i:I: OPT; do
	case "$OPT" in
		'd')
			DEVICE_FOUND=0
			while read \
				MAJOR MINOR DEVICE_NAME \
				READS_COMPLETED READS_MERGED SECTORS_READ TIME_SPENT_READING \
				WRITES_COMPLETED WRITES_MERGED SECTORS_WRITTEN TIME_SPENT_WRITING \
				IO_IN_PROGRESS \
				TIME_SPENT_IO TIME_SPENT_IO_WEIGHTED \
				DISCARDS_MERGED SECTORS_DISCARDED TIME_SPENT_DISCARDING \
				FLUSH_REQUESTS TIME_SPENT_FLUSHING \
				FUTURE_THINGIES
			do
				case "$DEVICE_NAME" in
					$OPTARG)
						DEVICE_FOUND=1
						# Checking the thresholds
						if [ -n "$THRESHOLD_IO_IN_PROGRESS_CRITICAL" ] && [ "$IO_IN_PROGRESS" -gt "$THRESHOLD_IO_IN_PROGRESS_CRITICAL" ]; then
							OUTPUT_EXIT_STATUS=$STATE_CRITICAL
							OUTPUT_DETAIL_CRITICAL="$OUTPUT_DETAIL_CRITICAL I/O in progress on $DEVICE_NAME: $IO_IN_PROGRESS"
						fi
						if [ -n "$THRESHOLD_IO_IN_PROGRESS_WARNING" ] && [ "$OUTPUT_EXIT_STATUS" != "$STATE_CRITICAL" ] && [ "$IO_IN_PROGRESS" -gt "$THRESHOLD_IO_IN_PROGRESS_WARNING" ]; then
							OUTPUT_EXIT_STATUS=$STATE_WARNING
							OUTPUT_DETAIL_WARNING="$OUTPUT_DETAIL_WARNING I/O in progress on $DEVICE_NAME: $IO_IN_PROGRESS"
						fi
						# Generating performance data
						# reads
						OUTPUT_PERFDATA="$( printf "%s\n'%s_reads_completed'=%dc;;;0" "$OUTPUT_PERFDATA" "$DEVICE_NAME" "$READS_COMPLETED" )"
						OUTPUT_PERFDATA="$( printf "%s\n'%s_reads_merged'=%dc;;;0" "$OUTPUT_PERFDATA" "$DEVICE_NAME" "$READS_MERGED" )"
						OUTPUT_PERFDATA="$( printf "%s\n'%s_sectors_read'=%dc;;;0" "$OUTPUT_PERFDATA" "$DEVICE_NAME" "$SECTORS_READ" )"
						OUTPUT_PERFDATA="$( printf "%s\n'%s_time_spent_reading'=%dc;;;0" "$OUTPUT_PERFDATA" "$DEVICE_NAME" "$TIME_SPENT_READING" )"
						# writes
						OUTPUT_PERFDATA="$( printf "%s\n'%s_writes_completed'=%dc;;;0" "$OUTPUT_PERFDATA" "$DEVICE_NAME" "$WRITES_COMPLETED" )"
						OUTPUT_PERFDATA="$( printf "%s\n'%s_writes_merged'=%dc;;;0" "$OUTPUT_PERFDATA" "$DEVICE_NAME" "$WRITES_MERGED" )"
						OUTPUT_PERFDATA="$( printf "%s\n'%s_sectors_written'=%dc;;;0" "$OUTPUT_PERFDATA" "$DEVICE_NAME" "$SECTORS_WRITTEN" )"
						OUTPUT_PERFDATA="$( printf "%s\n'%s_time_spent_writing'=%dc;;;0" "$OUTPUT_PERFDATA" "$DEVICE_NAME" "$TIME_SPENT_WRITING" )"
						# I/O
						OUTPUT_PERFDATA="$( printf "%s\n'%s_io_in_progress'=%d;%s;%s;0" "$OUTPUT_PERFDATA" "$DEVICE_NAME" "$IO_IN_PROGRESS" "$THRESHOLD_IO_IN_PROGRESS_WARNING" "$THRESHOLD_IO_IN_PROGRESS_CRITICAL" )"
						OUTPUT_PERFDATA="$( printf "%s\n'%s_time_spent_io'=%dc;;;0" "$OUTPUT_PERFDATA" "$DEVICE_NAME" "$TIME_SPENT_IO" )"
						OUTPUT_PERFDATA="$( printf "%s\n'%s_time_spent_io_weighted'=%dc;;;0" "$OUTPUT_PERFDATA" "$DEVICE_NAME" "$TIME_SPENT_IO_WEIGHTED" )"
						# Discards
						OUTPUT_PERFDATA="$( printf "%s\n'%s_discards_merged'=%dc;;;0" "$OUTPUT_PERFDATA" "$DEVICE_NAME" "$DISCARDS_MERGED" )"
						OUTPUT_PERFDATA="$( printf "%s\n'%s_sectors_discarded'=%dc;;;0" "$OUTPUT_PERFDATA" "$DEVICE_NAME" "$SECTORS_DISCARDED" )"
						OUTPUT_PERFDATA="$( printf "%s\n'%s_time_spent_discarding'=%dc;;;0" "$OUTPUT_PERFDATA" "$DEVICE_NAME" "$TIME_SPENT_DISCARDING" )"
						# Flushes
						OUTPUT_PERFDATA="$( printf "%s\n'%s_flush_requests'=%dc;;;0" "$OUTPUT_PERFDATA" "$DEVICE_NAME" "$FLUSH_REQUESTS" )"
						OUTPUT_PERFDATA="$( printf "%s\n'%s_time_spent_flushing'=%dc;;;0" "$OUTPUT_PERFDATA" "$DEVICE_NAME" "$TIME_SPENT_FLUSHING" )"
						;;
				esac
			done < /proc/diskstats
			# Check if the device requested has been found.
			if [ "$DEVICE_FOUND" -lt 1 ]; then
				echo "UNKNOWN device '$OPTARG' not found."
				exit $STATE_UNKNOWN
			fi
			;;

		'h')
			usage
			exit 0
			;;

		'i')
			THRESHOLD_IO_IN_PROGRESS_WARNING="$OPTARG"
			;;

		'I')
			THRESHOLD_IO_IN_PROGRESS_CRITICAL="$OPTARG"
			;;

		\?)
			usage
			exit 1
			;;
	esac
done

# Should we check if no device has been requested ? Overkill ?

# Final ouput
case "$OUTPUT_EXIT_STATUS" in
	'0')
		printf "OK%s" "$OUTPUT_DETAIL_OK"
		;;
	'1')
		printf "WARNING%s" "$OUTPUT_DETAIL_WARNING"
		;;
	'2')
		printf "CRITICAL%s" "$OUTPUT_DETAIL_CRITICAL"
		;;
	*)
		printf "UNKNOWN"
		;;
esac

# We sort the perfdata because of a weird bug in some icinga/pnp4nagios interactions
printf "|%s\n" "$( printf "%s" "$OUTPUT_PERFDATA" | sort | grep -v "^$" | tr "\n" " " )"
exit $OUTPUT_EXIT_STATUS