#!/usr/bin/env bash
set -euo pipefail

# This check_hpraid is a fork from check_cciss v0.15 written by Simone Rosa.
# Usage of old tools and drivers were removed to use only the smartpqi driver and the ssacli tool from HP.
# Upstream now at:
# https://gitea.evolix.org/evolix/ansible-roles/src/branch/stable/nagios-nrpe/files/plugins
# Source of the fork:
# https://exchange.nagios.org/directory/Plugins/Hardware/Storage-Systems/RAID-Controllers/check_cciss--2D-HP-and-Compaq-Smart-Array-Hardware-status/details
#
# Licence: GPLv2
# Description:
#
# This plugin checks hardware status for Smart Array Controllers,
# using HPE Smart Storage Administrator. It should support Debian 9 and over.
# (Array, controller, cache, battery, etc...)
#
# NOTE:
#
# You need to install the proprietary tool HPE Smart Storage Administrator (ssacli) from: 
# https://downloads.linux.hpe.com/SDR/repo/mcp
# Also NRPE need to launch ssacli as root.
#
# Please add this line to /etc/sudoers :
# --------------------------------------------------
# nagios      ALL=NOPASSWD: /usr/sbin/ssacli
#
# Examples:
#
#   ./check_cciss
# ----------------
# RAID OK
#
#   ./check_cciss -v
# -------------------
# RAID OK:  Smart Array 6i in Slot 0 array A logicaldrive 1 (67.8 GB, RAID 1+0, OK)
#           [Controller Status: OK  Cache Status: OK  Battery Status: OK]
#
# RAID CRITICAL - HP Smart Array Failed:  Smart Array 6i in Slot 0 (Embedded) \
#          array A logicaldrive 1 (33.9 GB, RAID 1, Interim Recovery Mode) \
#          physicaldrive 1:0 (port 1:id 0 , Parallel SCSI, --- GB, Failed)
#
# RAID WARNING - HP Smart Array Rebuilding:  Smart Array 6i in Slot 0 (Embedded) \
#          array A logicaldrive 1 (33.9 GB, RAID 1, Recovering, 26% complete) \
#          physicaldrive 1:0 (port 1:id 0 , Parallel SCSI, 36.4 GB, Rebuilding)
#
# ./check_cciss -v -p
# --------------------
# RAID OK:  Smart Array 6i in Slot 0 (Embedded) array A logicaldrive 1 (33.9 GB, RAID 1, OK)
#           physicaldrive 2:0 (port 2:id 0 , Parallel SCSI, 36.4 GB, OK)
#           physicaldrive 2:1 (port 2:id 1 , Parallel SCSI, 36.4 GB, OK)
#           physicaldrive 1:5 (port 1:id 5 , Parallel SCSI, 72.8 GB, OK, spare)
#           [Controller Status: OK Cache Status: OK Battery/Capacitor Status: OK]
#
# RAID CRITICAL - HP Smart Array Failed:  Smart Array 6i in Slot 0 (Embedded) \
#          array A logicaldrive 1 (33.9 GB, RAID 1, Interim Recovery Mode) \
#          physicaldrive 1:0 (port 1:id 0 , Parallel SCSI, --- GB, Failed) \
#          physicaldrive 1:1 (port 1:id 1 , Parallel SCSI, 36.4 GB, OK)
#
# RAID WARNING - HP Smart Array Rebuilding:  Smart Array 6i in Slot 0 (Embedded) \
#          array A logicaldrive 1 (33.9 GB, RAID 1, Recovering, 26% complete) \
#          physicaldrive 1:0 (port 1:id 0 , Parallel SCSI, 36.4 GB, Rebuilding) \
#          physicaldrive 1:1 (port 1:id 1 , Parallel SCSI, 36.4 GB, OK)
#
# ./check_cciss -v -b
# ----------------
#
# RAID OK:  Smart Array 6i in Slot 0 (Embedded) array A logicaldrive 1 (33.9 GB, RAID 1, OK) [Controller Status: OK]
#
#  [insted of]
# RAID CRITICAL - HP Smart Array Failed:  Smart Array 6i in Slot 0 (Embedded) \
#                 Controller Status: OK Cache Status: Temporarily Disabled \
#                 Battery/Capacitor Status: Failed (Replace Batteries/Capacitors)

PROGNAME=$(basename $0)
NAGIOS_PLUGINS="/usr/lib/nagios/plugins/"
REVISION=$(echo '0.16-evolix')
DEBUG="0"
VERBOSE="0"
ssacli=$(command -v ssacli)
PHYSICAL_DRIVE=0

. ${NAGIOS_PLUGINS}/utils.sh

print_usage() {
    echo ""
    echo "Usage: $PROGNAME [-v] [-p] [-e <number>] [-E <name>] [-b] [-s] [-d]"
    echo "Usage: $PROGNAME [-h]"
    echo "Usage: $PROGNAME [-V]"
    echo ""
    echo "  -v                   = show status and informations about RAID"
    echo "  -p                   = show detail for physical drives"
    echo "  -e <number>          = exclude slot number"
    echo "  -b                   = exclude battery/capacitor/cache status check"
    echo "  -d                   = use for debug (command line mode)"
    echo "  -h                   = help information"
    echo "  -V                   = version information"
    echo ""
    echo " ============="
}

print_help() {
    print_revision $PROGNAME $REVISION
    echo ""
    print_usage
    echo ""
    echo "This plugin checks hardware status for Smart Array Controllers,"
    echo "using HPE Smart Storage Administrator."
    echo ""
    support
    exit 0
}

while getopts "N:cvpbsde:Vh" options
do
    case $options in
      N)  ;;
      c)  ;;
      v)  VERBOSE=1;;
      p)  PHYSICAL_DRIVE=1;;
      d)  DEBUG=1;;
      e)  EXCLUDE_SLOT=1
          excludeslot="$OPTARG";;
      b)  EXCLUDE_BATTERY=1;;
      V)  print_revision $PROGNAME $REVISION
          exit 0;;
      h)  print_help
          exit 0;;
      \?) print_usage
          exit 0;;
      *)  print_usage
          exit 0;;
    esac
done

# Use smartpqi driver
# https://manpages.debian.org/buster/manpages/smartpqi.4.en.html
if [ -d /sys/bus/pci/drivers/smartpqi ]; then
    DRIVER="/sys/bus/pci/drivers/smartpqi"
    driverPresent='YES.'
else
    driverPresent='No!'
fi
if [ "$DEBUG" = "1" ]; then
    echo "### Check if \"HP Smart Array\" ($DRIVER) is present >>>\n"${driverPresent}"\n"
fi
if [[ "$driverPresent" == "No!" ]]; then
    echo "RAID UNKNOWN - HP Smart Array not found"
    exit $STATE_UNKNOWN
fi

# Check if "HP Array Utility CLI" is present
if [ "$DEBUG" = "1" ]; then
    echo "### Check if \"ssacli\" is present >>>\n"
fi
if [ ! -x $ssacli ]; then
    if [ -x $ssacli ]; then
        if [ "$DEBUG" = "1" ]; then
            echo "### \"ssacli\" is present >>>\n"
        fi
    else
        echo "ERROR: ssacli tools should be installed and with right sudoers/permissions (see the notes above)"
        exit $STATE_UNKNOWN
    fi
fi

# Check if "HP Controller" work correctly
check=$(sudo -u root $ssacli controller all show status 2>&1)
status=$?
if [ "$DEBUG" = "1" ]; then
    echo "### Check if \"HP Controller\" work correctly >>>\n"${check}"\n"
fi
if test ${status} -ne 0; then
    echo "RAID UNKNOWN - $ssacli did not execute properly : "${check}
    exit $STATE_UNKNOWN
fi

# Get "Slot" & exclude slot needed
EXCLUDE_SLOT=${EXCLUDE_SLOT:-0}
if [ "$EXCLUDE_SLOT" = "1" ]; then
    slots=$(echo ${check} | egrep -o "Slot \w" | awk '{print $NF}' | grep -v "$excludeslot")
else
    slots=$(echo ${check} | egrep -o "Slot \w" | awk '{print $NF}')
fi
if [ "$DEBUG" = "1" ]; then
    echo "### Get \"Slot\" & exclude slot not needed >>>\n"${slots}"\n"
fi
for slot in $slots; do
    # Get "logicaldrive" for slot
    check2b=$(sudo -u root $ssacli controller slot=$slot logicaldrive all show 2>&1)
    status=$?
    if test ${status} -ne 0; then
        echo "RAID UNKNOWN - $ssacli did not execute properly : "${check2b}
        exit $STATE_UNKNOWN
    fi
    check2=${check2:-}
    check2="$check2$check2b"
    if [ "$DEBUG" = "1" ]; then
        echo "### Get \"logicaldrive\" for slot >>>\n"${check2b}"\n"
    fi

    # Get "physicaldrive" for slot
    if [ "$PHYSICAL_DRIVE" = "1" -o "$DEBUG" = "1" ]; then
        check2b=$(sudo -u root $ssacli controller slot=$slot physicaldrive all show | sed -e 's/\?/\-/g' 2>&1 | grep "physicaldrive")
    else
        check2b=$(sudo -u root $ssacli controller slot=$slot physicaldrive all show | sed -e 's/\?/\-/g' 2>&1 | grep "physicaldrive" | (grep "\(Failure\|Failed\|Rebuilding\)" || true))
    fi
    status=$?
    if [ "$PHYSICAL_DRIVE" = "1" -o "$DEBUG" = "1" ]; then
        if test ${status} -ne 0; then
            echo "RAID UNKNOWN - $ssacli did not execute properly : "${check2b}
            exit $STATE_UNKNOWN
        fi
    fi
    check2="$check2$check2b"
    if [ "$DEBUG" = "1" ]; then
        echo "### Get \"physicaldrive\" for slot >>>\n"${check2b}"\n"
    fi
done

# Check STATUS
if [ "$DEBUG" = "1" ]; then
    echo "### Check STATUS >>>"
fi

# Omit battery/capacitor/cache status check if requested
EXCLUDE_BATTERY=${EXCLUDE_BATTERY:-0}
if [ "$EXCLUDE_BATTERY" = "1" ]; then
    check=$(echo "$check" | grep -v 'Battery/Capacitor Status: Failed (Replace Batteries/Capacitors)')
    check=$(echo "$check" | grep -v 'Cache Status: Temporarily Disabled')
fi

if echo ${check} | egrep Failed >/dev/null; then
    echo "RAID CRITICAL - HP Smart Array Failed: "${check} | egrep Failed
    exit $STATE_CRITICAL
elif echo ${check} | egrep Disabled >/dev/null; then
    echo "RAID CRITICAL - HP Smart Array Problem: "${check} | egrep Disabled
    exit $STATE_CRITICAL
elif echo ${check2} | egrep Failed >/dev/null; then
    echo "RAID CRITICAL - HP Smart Array Failed: "${check2} | egrep Failed
    exit $STATE_CRITICAL
elif echo ${check2} | egrep Failure >/dev/null; then
    echo "RAID WARNING - Component Failure: "${check2} | egrep Failure
    exit $STATE_WARNING
elif echo ${check2} | egrep Rebuild >/dev/null; then
    echo "RAID WARNING - HP Smart Array Rebuilding: "${check2} | egrep Rebuild
    exit $STATE_WARNING
elif echo ${check2} | egrep Recover >/dev/null; then
    echo "RAID WARNING - HP Smart Array Recovering: "${check2} | egrep Recover
    exit $STATE_WARNING
elif echo ${check} | egrep "Cache Status: Temporarily Disabled" >/dev/null; then
    echo "RAID WARNING - HP Smart Array Cache Disabled: "${check}
    exit $STATE_WARNING
elif echo ${check} | egrep FIRMWARE >/dev/null; then
    echo "RAID WARNING - "${check}
    exit $STATE_WARNING
else
    if [ "$DEBUG" = "1" -o "$VERBOSE" = "1" ]; then
        check3=$(echo "${check}" | egrep Status)
        check3=$(echo ${check3})
        echo "RAID OK: "${check2}" ["${check3}"]"
    else
        echo "RAID OK"
    fi
    exit $STATE_OK
fi

exit $STATE_UNKNOWN