textfile example script rework (#1074)
* textfile smartmon.sh Added functions to also parse megaraid disks. Added parsing to also detect the grown_defects counters. * textfile storcli.py Reworked the example file to export lots more information about megaraid attached controllers, VDs and PDs. Signed-off-by: Christopher Blum <christopher.blum@profitbricks.com>
This commit is contained in:
Christopher Blum
Ben Kochie
parent
1c9ea46cca
commit
6aa5cfba6c
2 changed files with 227 additions and 143 deletions
|
|
@ -7,7 +7,11 @@
|
|||
# data in them than you'd think.
|
||||
# http://arstechnica.com/civis/viewtopic.php?p=22062211
|
||||
|
||||
parse_smartctl_attributes_awk="$(cat << 'SMARTCTLAWK'
|
||||
# Formatting done via shfmt -i 2
|
||||
# https://github.com/mvdan/sh
|
||||
|
||||
parse_smartctl_attributes_awk="$(
|
||||
cat <<'SMARTCTLAWK'
|
||||
$1 ~ /^ *[0-9]+$/ && $2 ~ /^[a-zA-Z0-9_-]+$/ {
|
||||
gsub(/-/, "_");
|
||||
printf "%s_value{%s,smart_id=\"%s\"} %d\n", $2, labels, $1, $4
|
||||
|
|
@ -18,7 +22,8 @@ $1 ~ /^ *[0-9]+$/ && $2 ~ /^[a-zA-Z0-9_-]+$/ {
|
|||
SMARTCTLAWK
|
||||
)"
|
||||
|
||||
smartmon_attrs="$(cat << 'SMARTMONATTRS'
|
||||
smartmon_attrs="$(
|
||||
cat <<'SMARTMONATTRS'
|
||||
airflow_temperature_cel
|
||||
command_timeout
|
||||
current_pending_sector
|
||||
|
|
@ -64,63 +69,65 @@ parse_smartctl_attributes() {
|
|||
local disk_type="$2"
|
||||
local labels="disk=\"${disk}\",type=\"${disk_type}\""
|
||||
local vars="$(echo "${smartmon_attrs}" | xargs | tr ' ' '|')"
|
||||
sed 's/^ \+//g' \
|
||||
| awk -v labels="${labels}" "${parse_smartctl_attributes_awk}" 2>/dev/null \
|
||||
| tr A-Z a-z \
|
||||
| grep -E "(${smartmon_attrs})"
|
||||
sed 's/^ \+//g' |
|
||||
awk -v labels="${labels}" "${parse_smartctl_attributes_awk}" 2>/dev/null |
|
||||
tr A-Z a-z |
|
||||
grep -E "(${smartmon_attrs})"
|
||||
}
|
||||
|
||||
parse_smartctl_scsi_attributes() {
|
||||
local disk="$1"
|
||||
local disk_type="$2"
|
||||
local labels="disk=\"${disk}\",type=\"${disk_type}\""
|
||||
while read line ; do
|
||||
attr_type="$(echo "${line}" | tr '=' ':' | cut -f1 -d: | sed 's/^ \+//g' | tr ' ' '_')"
|
||||
attr_value="$(echo "${line}" | tr '=' ':' | cut -f2 -d: | sed 's/^ \+//g')"
|
||||
case "${attr_type}" in
|
||||
number_of_hours_powered_up_) power_on="$( echo "${attr_value}" | awk '{ printf "%e\n", $1 }')" ;;
|
||||
Current_Drive_Temperature) temp_cel="$(echo ${attr_value} | cut -f1 -d' ' | awk '{ printf "%e\n", $1 }')" ;;
|
||||
Blocks_read_from_cache_and_sent_to_initiator_) lbas_read="$(echo ${attr_value} | awk '{ printf "%e\n", $1 }')" ;;
|
||||
Accumulated_start-stop_cycles) power_cycle="$(echo ${attr_value} | awk '{ printf "%e\n", $1 }')" ;;
|
||||
esac
|
||||
done
|
||||
echo "power_on_hours_raw_value{"${labels}",smart_id=\"9\"} ${power_on}"
|
||||
echo "temperature_celsius_raw_value{"${labels}",smart_id=\"194\"} ${temp_cel}"
|
||||
echo "total_lbas_read_raw_value{"${labels}",smart_id=\"242\"} ${lbas_read}"
|
||||
echo "power_cycle_count_raw_value{"${labels}",smart_id=\"12\"} ${power_cycle}"
|
||||
local disk="$1"
|
||||
local disk_type="$2"
|
||||
local labels="disk=\"${disk}\",type=\"${disk_type}\""
|
||||
while read line; do
|
||||
attr_type="$(echo "${line}" | tr '=' ':' | cut -f1 -d: | sed 's/^ \+//g' | tr ' ' '_')"
|
||||
attr_value="$(echo "${line}" | tr '=' ':' | cut -f2 -d: | sed 's/^ \+//g')"
|
||||
case "${attr_type}" in
|
||||
number_of_hours_powered_up_) power_on="$(echo "${attr_value}" | awk '{ printf "%e\n", $1 }')" ;;
|
||||
Current_Drive_Temperature) temp_cel="$(echo ${attr_value} | cut -f1 -d' ' | awk '{ printf "%e\n", $1 }')" ;;
|
||||
Blocks_read_from_cache_and_sent_to_initiator_) lbas_read="$(echo ${attr_value} | awk '{ printf "%e\n", $1 }')" ;;
|
||||
Accumulated_start-stop_cycles) power_cycle="$(echo ${attr_value} | awk '{ printf "%e\n", $1 }')" ;;
|
||||
Elements_in_grown_defect_list) grown_defects="$(echo ${attr_value} | awk '{ printf "%e\n", $1 }')" ;;
|
||||
esac
|
||||
done
|
||||
[ ! -z "$power_on" ] && echo "power_on_hours_raw_value{${labels},smart_id=\"9\"} ${power_on}"
|
||||
[ ! -z "$temp_cel" ] && echo "temperature_celsius_raw_value{${labels},smart_id=\"194\"} ${temp_cel}"
|
||||
[ ! -z "$lbas_read" ] && echo "total_lbas_read_raw_value{${labels},smart_id=\"242\"} ${lbas_read}"
|
||||
[ ! -z "$power_cycle" ] && echo "power_cycle_count_raw_value{${labels},smart_id=\"12\"} ${power_cycle}"
|
||||
[ ! -z "$grown_defects" ] && echo "grown_defects_count_raw_value{${labels},smart_id=\"12\"} ${grown_defects}"
|
||||
}
|
||||
|
||||
parse_smartctl_info() {
|
||||
local -i smart_available=0 smart_enabled=0 smart_healthy=0
|
||||
local disk="$1" disk_type="$2"
|
||||
local model_family='' device_model='' serial_number='' fw_version='' vendor='' product='' revision='' lun_id=''
|
||||
while read line ; do
|
||||
while read line; do
|
||||
info_type="$(echo "${line}" | cut -f1 -d: | tr ' ' '_')"
|
||||
info_value="$(echo "${line}" | cut -f2- -d: | sed 's/^ \+//g' | sed 's/"/\\"/')"
|
||||
case "${info_type}" in
|
||||
Model_Family) model_family="${info_value}" ;;
|
||||
Device_Model) device_model="${info_value}" ;;
|
||||
Serial_Number) serial_number="${info_value}" ;;
|
||||
Firmware_Version) fw_version="${info_value}" ;;
|
||||
Vendor) vendor="${info_value}" ;;
|
||||
Product) product="${info_value}" ;;
|
||||
Revision) revision="${info_value}" ;;
|
||||
Logical_Unit_id) lun_id="${info_value}" ;;
|
||||
Model_Family) model_family="${info_value}" ;;
|
||||
Device_Model) device_model="${info_value}" ;;
|
||||
Serial_Number) serial_number="${info_value}" ;;
|
||||
Firmware_Version) fw_version="${info_value}" ;;
|
||||
Vendor) vendor="${info_value}" ;;
|
||||
Product) product="${info_value}" ;;
|
||||
Revision) revision="${info_value}" ;;
|
||||
Logical_Unit_id) lun_id="${info_value}" ;;
|
||||
esac
|
||||
if [[ "${info_type}" == 'SMART_support_is' ]] ; then
|
||||
if [[ "${info_type}" == 'SMART_support_is' ]]; then
|
||||
case "${info_value:0:7}" in
|
||||
Enabled) smart_enabled=1 ;;
|
||||
Availab) smart_available=1 ;;
|
||||
Unavail) smart_available=0 ;;
|
||||
Enabled) smart_enabled=1 ;;
|
||||
Availab) smart_available=1 ;;
|
||||
Unavail) smart_available=0 ;;
|
||||
esac
|
||||
fi
|
||||
if [[ "${info_type}" == 'SMART_overall-health_self-assessment_test_result' ]] ; then
|
||||
if [[ "${info_type}" == 'SMART_overall-health_self-assessment_test_result' ]]; then
|
||||
case "${info_value:0:6}" in
|
||||
PASSED) smart_healthy=1 ;;
|
||||
PASSED) smart_healthy=1 ;;
|
||||
esac
|
||||
elif [[ "${info_type}" == 'SMART_Health_Status' ]] ; then
|
||||
elif [[ "${info_type}" == 'SMART_Health_Status' ]]; then
|
||||
case "${info_value:0:2}" in
|
||||
OK) smart_healthy=1 ;;
|
||||
OK) smart_healthy=1 ;;
|
||||
esac
|
||||
fi
|
||||
done
|
||||
|
|
@ -130,7 +137,8 @@ parse_smartctl_info() {
|
|||
echo "device_smart_healthy{disk=\"${disk}\",type=\"${disk_type}\"} ${smart_healthy}"
|
||||
}
|
||||
|
||||
output_format_awk="$(cat << 'OUTPUTAWK'
|
||||
output_format_awk="$(
|
||||
cat <<'OUTPUTAWK'
|
||||
BEGIN { v = "" }
|
||||
v != $1 {
|
||||
print "# HELP smartmon_" $1 " SMART metric " $1;
|
||||
|
|
@ -142,15 +150,15 @@ OUTPUTAWK
|
|||
)"
|
||||
|
||||
format_output() {
|
||||
sort \
|
||||
| awk -F'{' "${output_format_awk}"
|
||||
sort |
|
||||
awk -F'{' "${output_format_awk}"
|
||||
}
|
||||
|
||||
smartctl_version="$(/usr/sbin/smartctl -V | head -n1 | awk '$1 == "smartctl" {print $2}')"
|
||||
smartctl_version="$(/usr/sbin/smartctl -V | head -n1 | awk '$1 == "smartctl" {print $2}')"
|
||||
|
||||
echo "smartctl_version{version=\"${smartctl_version}\"} 1" | format_output
|
||||
|
||||
if [[ "$(expr "${smartctl_version}" : '\([0-9]*\)\..*')" -lt 6 ]] ; then
|
||||
if [[ "$(expr "${smartctl_version}" : '\([0-9]*\)\..*')" -lt 6 ]]; then
|
||||
exit
|
||||
fi
|
||||
|
||||
|
|
@ -159,13 +167,18 @@ device_list="$(/usr/sbin/smartctl --scan-open | awk '/^\/dev/{print $1 "|" $3}')
|
|||
for device in ${device_list}; do
|
||||
disk="$(echo ${device} | cut -f1 -d'|')"
|
||||
type="$(echo ${device} | cut -f2 -d'|')"
|
||||
echo "smartctl_run{disk=\"${disk}\",type=\"${type}\"}" $(TZ=UTC date '+%s')
|
||||
echo "smartctl_run{disk=\"${disk}\",type=\"${type}\"}" "$(TZ=UTC date '+%s')"
|
||||
# Get the SMART information and health
|
||||
/usr/sbin/smartctl -i -H -d "${type}" "${disk}" | parse_smartctl_info "${disk}" "${type}"
|
||||
# Get the SMART attributes
|
||||
case ${type} in
|
||||
sat) /usr/sbin/smartctl -A -d "${type}" "${disk}" | parse_smartctl_attributes "${disk}" "${type}" ;;
|
||||
scsi) /usr/sbin/smartctl -A -d "${type}" "${disk}" | parse_smartctl_scsi_attributes "${disk}" "${type}" ;;
|
||||
*) echo "disk type is not sat or scsi, ${type}"; exit ;;
|
||||
sat) /usr/sbin/smartctl -A -d "${type}" "${disk}" | parse_smartctl_attributes "${disk}" "${type}" ;;
|
||||
sat+megaraid*) /usr/sbin/smartctl -A -d "${type}" "${disk}" | parse_smartctl_attributes "${disk}" "${type}" ;;
|
||||
scsi) /usr/sbin/smartctl -A -d "${type}" "${disk}" | parse_smartctl_scsi_attributes "${disk}" "${type}" ;;
|
||||
megaraid*) /usr/sbin/smartctl -A -d "${type}" "${disk}" | parse_smartctl_scsi_attributes "${disk}" "${type}" ;;
|
||||
*)
|
||||
echo "disk type is not sat, scsi or megaraid but ${type}"
|
||||
exit
|
||||
;;
|
||||
esac
|
||||
done | format_output
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue