scripts/xymon/plugins/client/ext/smart

153 lines
4.2 KiB
Plaintext
Raw Normal View History

#!/bin/sh
# NOTE: Must be run as root, so you probably need to setup sudo for this.
# Vars {{{
debug="0"
## Colors {{{
c_redb='\033[1;31m'
c_magentab='\033[1;35m'
c_reset='\033[0m'
## }}}
# }}}
# Functions
## Test if a disk really support SMART {{{
## Smartctl can give an health status even without a full support
## of SMART for some type (eg. scsi or megaraid).
## Exemple: SMART support is: Unavailable - device lacks SMART capability.
is_disk_support_smart() {
_disk="${1}"
_type="${2}"
_smarctl_support_result="/tmp/dsupport.$(basename "${_disk}")"
smart_support_msg=""
[ "${debug}" -eq "0" ] && printf "${c_magentab}%-6b${c_reset}\n" "DEBUG: is_disk_support_smart func check if SMART is supported on: ${_disk}."
if test -f "${_smarctl_support_result}"; then rm -f -- "${_smarctl_support_result}"; fi
## Grep only "support" lines from disk's informations
smartctl -d "${_type}" -i -- "${_disk}" | grep -E "^SMART support is:" -- > "${_smarctl_support_result}"
## If the file is not empty
if test -s "${_smarctl_support_result}"; then
## Parse all "support" lines
while IFS= read -r _LINE; do
if ! printf -- '%s' "${_LINE}" | grep -q -E -- "(Enabled|Available)"
then
smart_support_msg="${_LINE}"
fi
done < "${_smarctl_support_result}"
else
smart_support_msg="smartctl was not able to open ${_disk} DEVICE with ${_type} TYPE."
fi
if [ -z "${smart_support_msg}" ]; then
[ "${debug}" -eq "0" ] && printf "${c_magentab}%-6b${c_reset}\n" "DEBUG: is_disk_support_smart func SMART seems fully supported on: ${_disk} with ${_type} type."
else
[ "${debug}" -eq "0" ] && printf "${c_magentab}%-6b${c_reset}\n" "DEBUG: is_disk_support_smart func SMART is not fully supported on: ${_disk} with ${_type} type. See smartctl informations:\n${smart_support_msg}"
fi
rm -f -- "${_smarctl_support_result}"
}
## }}}
## Test the type of disk with smartctl {{{
choose_correct_type() {
_disk="${1}"
_scanned_type="${2}"
_default_type="auto"
TYPE=""
SMART_SUPPORT_MSG=""
for test_type in "${_default_type}" "${_scanned_type}"; do
is_disk_support_smart "${_disk}" "${test_type}"
## If no message, the type is correct
if [ -z "${smart_support_msg}" ]; then
TYPE="${test_type}"
SMART_SUPPORT_MSG=""
return
else
SMART_SUPPORT_MSG="${smart_support_msg}"
fi
done
}
## }}}
if test -f /tmp/dres; then rm -f -- /tmp/dres; fi
if test -f /tmp/dscan; then rm -f -- /tmp/dscan; fi
2020-02-19 16:00:36 +01:00
# Get the list of all available devices
2020-02-19 11:18:25 +01:00
smartctl --scan > /tmp/dscan
# TODO: Test if the file is not empty
2020-02-19 16:00:36 +01:00
while IFS= read -r LINE; do
## Get device path
2020-02-19 11:18:25 +01:00
DISK=$(echo "${LINE}" | cut -d" " -f1)
## Try to determine the best type
SCANNED_TYPE=$(echo "${LINE}" | cut -d" " -f3)
choose_correct_type "${DISK}" "${SCANNED_TYPE}"
## If no correct type was found for this device
if [ -z "${TYPE}" ]; then
[ "${debug}" -eq "0" ] && printf "${c_magentab}%-6b${c_reset}\n" "DEBUG: SMART is not fully supported."
DRES=$(printf '%s' "${SMART_SUPPORT_MSG}")
DCODE="2"
else
[ "${debug}" -eq "0" ] && printf "${c_magentab}%-6b${c_reset}\n" "DEBUG: SMART seems fully supported, proceed normally."
### Get SMART Health Status and return code
DRES=$(/usr/sbin/smartctl -H -d "${TYPE}" -n standby "${DISK}")
DCODE=$?
fi
DSTBY=$(( DCODE & 2 ))
DFAIL=$(( DCODE & 8 ))
DWARN=$(( DCODE & 32 ))
2020-02-19 16:00:36 +01:00
## Give a weight to each to color to easily get the page status
if test $DSTBY -ne 0
then
COLOR="4&clear"
elif test $DFAIL -ne 0
then
COLOR="1&red"
elif test $DWARN -ne 0
then
COLOR="2&yellow"
else
COLOR="3&green"
fi
2020-02-19 11:18:25 +01:00
echo "${COLOR} $DISK ${TYPE}"
2020-02-19 11:18:25 +01:00
echo "${COLOR} $DISK ${TYPE}" | cut -c2- >>/tmp/dres
echo "" >>/tmp/dres
2020-02-19 16:00:36 +01:00
echo "$DRES" | grep -v -E "^smartctl|^Copyright|^$|^===" >>/tmp/dres
echo "------------------------------------------------------------" >>/tmp/dres
echo "" >>/tmp/dres
echo "" >>/tmp/dres
2020-02-19 11:18:25 +01:00
done < /tmp/dscan >/tmp/dcheck
2020-02-19 16:00:36 +01:00
# Set the global color according to the highest alert
COLOR=$(< /tmp/dcheck awk '{print $1}' | sort | uniq | head -1 | cut -c3-)
2020-02-19 16:00:36 +01:00
$XYMON "${XYMSRV}" "status ${MACHINE}.smart ${COLOR} SMART health check
$(< /tmp/dcheck cut -c2-)
==================== Detailed status ====================
$(cat /tmp/dres)
"
rm -f -- /tmp/dres /tmp/dcheck /tmp/dscan
exit 0