check_mdstat.sh

@@ -1,204 +1,47 @@
 #!/bin/bash
 #
 # Created by Sebastian Grewe, Jammicron Technology
-# Changes by Jasem Elayeb on 02.03.2016
+# Changes By Jasem Elayeb on 02.03.2016
-# Chagnes by Jonny007-MKD on 06.02.2020
 # JE: add Physical Disks Name RAID_DISKS
 # JE: add Physical Disks Status DISKS_STATUS
 # JE: add Array Names RAID_ARRAY
-# J007: Only check a single array
 
+# Get count of raid arrays
+RAID_DEVICES=`grep ^md -c /proc/mdstat`
 
-if [ $# -lt 1 -o -z "$1" ]; then
+# Get count of degraded arrays
-	echo "ERROR: pass raid name as argument"
+RAID_STATUS=`grep "\[.*_.*\]" /proc/mdstat -c`
-	exit 1;
-fi
-while [[ $# -gt 0 ]]; do
-	case $1 in
-		-r|--raid)	TARGET_RAID_NAME="$2";;
-		-?|--help)
-			echo "Check /proc/mdstat. Arguments:"
-			echo "--raid NAME: Raid name, e.g. md0"
-			;;
-	esac
-	shift
-done
 
-if [ -z "$TARGET_RAID_NAME" ]; then
+# Is an array currently recovering, get percentage of recovery
-	echo "ERROR: pass --raid as argument"
+RAID_RECOVER=`grep recovery /proc/mdstat | awk '{print $4}'`
-	exit 255
+
+# Is an array currently resyncing, get percentage of resync
+
+RAID_RESYNC=`grep resync /proc/mdstat | awk '{print $4}'`
+
+RAID_ARRAY=`awk '/md[1-9]/{for (i=1;i<=NF;++i) if ($i~/md[1-2]/)print $i}' /proc/mdstat |xargs`
+RAID_DISKS=`awk '/sd[a-z]/{for (i=1;i<=NF;++i) if ($i~/sd[a-z]/)print $i}' /proc/mdstat |xargs`
+DISKS_STATUS=`grep algorithm  /proc/mdstat|awk '{print $12}'`
+
+# Check raid status
+# RAID recovers --> Warning
+if [[ $RAID_RECOVER ]]; then
+        STATUS="WARNING - Checked $RAID_DEVICES arrays $RAID_ARRAY, recovering : $RAID_RECOVER"
+        EXIT=1
+# RAID resync --> Warning
+elif [[ $RAID_RESYNC ]]; then
+        STATUS="WARNING - Checked $RAID_DEVICES arrays $RAID_ARRAY., resyncing : $RAID_RESYNC"
+        EXIT=1
+# RAID ok
+elif [[ $RAID_STATUS  == "0" ]]; then
+        STATUS="OK - Checked $RAID_DEVICES arrays $RAID_ARRAY."
+        EXIT=0
+# All else critical, better save than sorry
+else
+        STATUS="CRITICAL - Checked $RAID_DEVICES arrays $RAID_ARRAY, $RAID_STATUS have FAILED"
+        EXIT=2
 fi
 
-function log {
+# Status and quit
-	local x=
+echo -e "$STATUS \n Physical Disks: $RAID_DISKS Disks Status: $DISKS_STATUS "
-	#echo "  > $@"
+exit $EXIT
-}
-
-CURRENT_RAID=	# string
-function setCurrentRaid {
-	# This function is called for each line in $1
-	# If the line does not start with a whitespace, it denotes a new RAID device and CURRENT_RAID is set
-	# Also, the state and the devices are parsed 
-	local line="$1"
-	if [[ $line =~ ^[^\s].+:.+ ]]; then
-		local x1="${line%:*}" # Remove : suffix
-		local x2="${x1% *}"   # Trim right
-		CURRENT_RAID="$x2"
-		log "SET CURRENT RAID"
-	fi
-	log "CURRENT RAID: $CURRENT_RAID"
-}
-
-STATE=		# "active", "started", "inactive"
-LEVEL=		# int
-DEVICES=	# strings separated by whitespace
-function setStateAndDevicesAndLevel {
-	# This functions is called for the line that starts a new RAID device by setCurrentRaid, with the complete line in $1
-	# It parses the STATE, the LEVEL and the DEVICES
-	local line="$1"
-	local x1="${line#*: }" 				# Remove : prefix
-	STATE="${x1%% *}" 					# Only the first word
-	local x2="${x1:$((${#STATE}+1))}"	# Remove the first word
-	LEVEL="${x2%% *}" 					# Only the first word
-	local x3="${x2:$((${#LEVEL}+1))}"	# Remove the first word
-
-	LEVEL="${LEVEL:4}"					# Remove the raid prefix
-
-	for dev1 in $x3; do
-		local dev2="${dev1%[*}"
-		if [ -n "$DEVICES" ]; then DEVICES="$DEVICES "; fi
-		DEVICES="${DEVICES}$dev2"
-	done
-
-	log "STATE = $STATE"
-	log "LEVEL = $LEVEL"
-	log "DEVICES = $DEVICES"
-}
-
-SIZE_IN_BLOCKS="!"	# int
-NUM_DEVICES=0		# int with number of devices in raid
-BAD_DEVICES=0		# int with number of "_" in "[UUU_UU]"
-function parseConfigStatusLine {
-	# This function is called for 1st line after the raid definition line
-	# It parses the SIZE_IN_BLOCKS
-	local line="$1"
-	line="${line:6}"			# trim left
-	SIZE_IN_BLOCKS=${line%% *}
-
-	local lastWord=${line##* }	# Get the last word
-	lastWord=${lastWord:1:-1}	# Remove first and last char
-	NUM_DEVICES=${#lastWord}	# str length = num devices
-	lastWord=${lastWord//U/}	# Remove all U, so only _ remain
-	BAD_DEVICES=${#lastWord}	# str length = num bad devices
-
-
-	log "SIZE_IN_BLOCKS = $SIZE_IN_BLOCKS"
-	log "NUM_DEVICES = $NUM_DEVICES"
-	log "BAD_DEVICES = $BAD_DEVICES"
-	# TODO
-}
-
-function parseBitmapOrRecoveryLine {
-	# This function is called for the 2nd line after the raid definition line
-	# It checks whether it contains a bitmap or a recovery line
-	if [[ "$1" == *bitmap* ]]; then
-		parseBitmapLine "$1"
-	fi
-	if [[ "$1" == *recovery* ]]; then
-		parseRecoveryLine "$1"
-	fi
-}
-
-function parseBitmapLine {
-	local x
-	# TODO
-}
-
-RECOVERY_PROGRESS=	# float in percent
-RECOVERY_INFO=		# finish and speed
-function parseRecoveryLine {
-	# This function is called for the 2nd or 3rd lineafter the raid definition line
-	# It sets the RECOVERY_PROGRESS
-	local line="$1"
-	
-	local x1="${line##*recovery = }"
-	local x2="${x1%%%*}"
-	RECOVERY_PROGRESS="$x2"
-	
-	local x3="${line##*finish=}"
-	RECOVERY_INFO="finish=$x3"
-
-	log "RECOVERY_PROGESS = $RECOVERY_PROGRESS"
-	log "RECOVERY_INFO = $RECOVERY_INFO"
-}
-
-function printOutputAndExit {
-	# STATE, LEVEL, DEVICES, SIZE_IN_BLOCKS, NUM_DEVICES, BAD_DEVICES, RECOVERY_PROGRESS, RECOVERY_INFO
-
-	local result="OK"
-	local info=""
-
-	if [ -n "$RECOVERY_PROGRESS" ]; then
-		result="WARNING"
-		if [ -n "$info" ]; then info="$info. "; fi
-		info="${info}Recovering: $RECOVERY_PROGRESS $RECOVERY_INFO"
-	fi
-
-	if [ "$STATE" != "active" -a "$STATE" != "started" ]; then
-		result="CRITICAL"
-		if [ -n "$info" ]; then info="$info. "; fi
-		info="${info}State is $STATE"
-	fi
-
-	if [ $BAD_DEVICES -gt 0 ]; then
-		result="CRITICAL"
-		if [ -n "$info" ]; then info="$info. "; fi
-		info="${info}Missing $NUM_DEVICES devices"
-	fi
-
-	echo -n "$result"
-	if [ -n "$info" ]; then
-		echo -n ": $info"
-	fi
-
-	echo -n " |"
-	echo -n " 'raid level'=$LEVEL"
-	echo -n " 'size in blocks'=$SIZE_IN_BLOCKS"
-	echo -n " 'num devices'=$NUM_DEVICES"
-	echo -n " 'num bad devices'=$BAD_DEVICES;;1;0;$NUM_DEVICES"
-	if [ -n "$RECOVERY_PROGRESS" ]; then
-		echo -n " 'recovery progress'=$RECOVERY_PROGRESS%;0.1;;0;100"
-	fi
-	echo
-
-	case $result in
-		OK)       exit 0;;
-		WARNING)  exit 1;;
-		CRITICAL) exit 2;;
-	esac
-}
-
-
-HAD_TARGET_RAID=0
-
-while IFS= read -r line; do
-    #echo "  : $line"
-
-	setCurrentRaid "$line"
-
-	if [ "$CURRENT_RAID" == "$TARGET_RAID_NAME" ]; then
-		case $HAD_TARGET_RAID in
-			0) setStateAndDevicesAndLevel "$line";;
-			1) parseConfigStatusLine "$line";;
-			2) parseBitmapOrRecoveryLine "$line";;
-			3) parseRecoveryLine "$line";;
-		esac
-		HAD_TARGET_RAID=$(($HAD_TARGET_RAID+1))
-	elif [ $HAD_TARGET_RAID -gt 0 ]; then
-		printOutputAndExit
-	fi
-done < /proc/mdstat
-
-echo "CRITICAL: RAID $TARGET_RAID_NAME not found in mdstat"
-exit 2
-

test/1missing.input

@@ -1,4 +0,0 @@
-Personalities : [raid6] [raid5] [raid4]
-md0 : active raid5 sda1[0] sdd1[2] sdb1[1]
-     1465151808 blocks level 5, 64k chunk, algorithm 2 [4/3] [UUU_]
-    unused devices: <none>

test/1missing.output

@@ -1 +0,0 @@
-CRITICAL: Missing 1 of 4 devices | 'raid level'=5 'size in blocks'=1465151808 'num devices'=4 'num bad devices'=1;;1;0;4

test/OK1.input

@@ -1,6 +0,0 @@
-Personalities : [raid1] 
-md0 : active raid1 sdb1[1] sda1[0]
-      976430080 blocks super 1.2 [2/2] [UU]
-      bitmap: 0/8 pages [0KB], 65536KB chunk
-
-unused devices: <none>

test/OK1.output

@@ -1 +0,0 @@
-OK | 'raid level'=1 'size in blocks'=976430080 'num devices'=2 'num bad devices'=0;;1;0;2

test/OK2.input

@@ -1,6 +0,0 @@
-Personalities : [raid1] [raid6] [raid5] [raid4]
-md0 : active raid5 sde1[0] sdf1[4] sdb1[5] sdd1[2] sdc1[1]
-      1250241792 blocks super 1.2 level 5, 64k chunk, algorithm 2 [5/5] [UUUUU]
-      bitmap: 0/10 pages [0KB], 16384KB chunk
-
-unused devices: <none>

test/OK2.output

@@ -1 +0,0 @@
-OK | 'raid level'=5 'size in blocks'=1250241792 'num devices'=5 'num bad devices'=0;;1;0;5

test/OK3.input

@@ -1,14 +0,0 @@
-Personalities : [raid1] [raid6] [raid5] [raid4]
-md1 : active raid1 sdb2[1] sda2[0]
-      136448 blocks [2/2] [UU]
-
-md2 : active raid1 sdb3[1] sda3[0]
-      129596288 blocks [2/2] [UU]
-
-md0 : active raid5 sdl1[9] sdk1[8] sdj1[7] sdi1[6] sdh1[5] sdg1[4] sdf1[3] sde1[2] sdd1[1] sdc1[0]
-      1318680576 blocks level 5, 1024k chunk, algorithm 2 [10/10] [UUUUUUUUUU]
-
-md3 : active raid1 sdb1[1] sda1[0]
-      16787776 blocks [2/2] [UU]
-
-unused devices: <none> 

test/OK3.output

@@ -1 +0,0 @@
-OK | 'raid level'=5 'size in blocks'=1318680576 'num devices'=10 'num bad devices'=0;;1;0;10

test/bitmap.input

@@ -1,7 +0,0 @@
-Personalities : [linear] [raid0] [raid1] [raid5] [raid4] [raid6]
-md0 : active raid6 sdf1[0] sde1[1] sdd1[2] sdc1[3] sdb1[4] sda1[5] hdb1[6]
-      1225557760 blocks level 6, 256k chunk, algorithm 2 [7/7] [UUUUUUU]
-      bitmap: 0/234 pages [0KB], 512KB chunk
-
-unused devices: <none>
-

test/bitmap.output

@@ -1 +0,0 @@
-OK | 'raid level'=6 'size in blocks'=1225557760 'num devices'=7 'num bad devices'=0;;1;0;7

test/notExisting.input

@@ -1,5 +0,0 @@
-Personalities : [raid1] [raid6] [raid5] [raid4]
-md1 : active raid1 sdb2[1] sda2[0]
-      136448 blocks [2/2] [UU]
-
-unused devices: <none> 

test/notExisting.output

@@ -1 +0,0 @@
-CRITICAL: RAID md0 not found in notExisting.input

test/recovery.input

@@ -1,6 +0,0 @@
-Personalities : [raid1] [raid6] [raid5] [raid4]
-md0 : active raid5 sdh1[6] sdg1[4] sdf1[3] sde1[2] sdd1[1] sdc1[0]
-      1464725760 blocks level 5, 64k chunk, algorithm 2 [6/5] [UUUUUU]
-      [==>..................]  recovery = 12.6% (37043392/292945152) finish=127.5min speed=33440K/sec
-
-unused devices: <none>

test/recovery.output

@@ -1 +0,0 @@
-WARNING: Recovering: progress=12.6% finish=127.5min speed=33440K/sec | 'raid level'=5 'size in blocks'=1464725760 'num devices'=6 'num bad devices'=0;;1;0;6 'recovery progress'=12.6%;0;;0;100

test/run.sh

@@ -1,22 +0,0 @@
-dir="`dirname $0`"
-
-ERROR=false
-for f in `ls $dir/*.input`; do
-	RESULT="`$dir/../check_mdstat.sh --raid md0 --input $f`"
-	EXPECTED="`cat $dir/$(basename $f .input).output`"
-	if [ "$RESULT" != "$EXPECTED" ]; then
-		echo "Error for test $(basename $f .input):"
-		echo "	Expected"
-		echo "		$EXPECTED"
-		echo "	and got"
-		echo "		$RESULT"
-		ERROR=true
-	fi
-done
-
-if $ERROR; then
-	exit 1;
-else
-	echo "Tests passed."
-	exit 0;
-fi