checks/supermicro

   1 #!/usr/bin/python
   2 # -*- encoding: utf-8; py-indent-offset: 4 -*-
   3 # +------------------------------------------------------------------+
   4 # |             ____ _               _        __  __ _  __           |
   5 # |            / ___| |__   ___  ___| | __   |  \/  | |/ /           |
   6 # |           | |   | '_ \ / _ \/ __| |/ /   | |\/| | ' /            |
   7 # |           | |___| | | |  __/ (__|   <    | |  | | . \            |
   8 # |            \____|_| |_|\___|\___|_|\_\___|_|  |_|_|\_\           |
   9 # |                                                                  |
  10 # | Copyright Mathias Kettner 2016             mk@mathias-kettner.de |
  11 # +------------------------------------------------------------------+
  12 #
  13 # This file is part of Check_MK.
  14 # The official homepage is at http://mathias-kettner.de/check_mk.
  15 #
  16 # check_mk is free software;  you can redistribute it and/or modify it
  17 # under the  terms of the  GNU General Public License  as published by
  18 # the Free Software Foundation in version 2.  check_mk is  distributed
  19 # in the hope that it will be useful, but WITHOUT ANY WARRANTY;  with-
  20 # out even the implied warranty of  MERCHANTABILITY  or  FITNESS FOR A
  21 # PARTICULAR PURPOSE. See the  GNU General Public License for more de-
  22 # tails. You should have  received  a copy of the  GNU  General Public
  23 # License along with GNU Make; see the file  COPYING.  If  not,  write
  24 # to the Free Software Foundation, Inc., 51 Franklin St,  Fifth Floor,
  25 # Boston, MA 02110-1301 USA.
  26
  27 # .1.3.6.1.4.1.10876.2.1.1.1.1.2.1 Fan1 Fan Speed
  28 # .1.3.6.1.4.1.10876.2.1.1.1.1.2.2 Fan2 Fan Speed
  29 # ...
  30 # .1.3.6.1.4.1.10876.2.1.1.1.1.2.6 Vcore Voltage
  31 # .1.3.6.1.4.1.10876.2.1.1.1.1.2.7 CPU VTT Voltage
  32 # .1.3.6.1.4.1.10876.2.1.1.1.1.3.1 0
  33 # .1.3.6.1.4.1.10876.2.1.1.1.1.3.2 0
  34 # ...
  35 # .1.3.6.1.4.1.10876.2.1.1.1.1.3.6 1
  36 # .1.3.6.1.4.1.10876.2.1.1.1.1.3.7 1
  37 # .1.3.6.1.4.1.10876.2.1.1.1.1.4.1 3760
  38 # .1.3.6.1.4.1.10876.2.1.1.1.1.4.2 1909
  39 # ...
  40 # .1.3.6.1.4.1.10876.2.1.1.1.1.4.6 1080
  41 # .1.3.6.1.4.1.10876.2.1.1.1.1.4.7 1056
  42 # ...
  43 # .1.3.6.1.4.1.10876.2.1.1.1.1.5.6 1488
  44 # .1.3.6.1.4.1.10876.2.1.1.1.1.5.7 1344
  45 # .1.3.6.1.4.1.10876.2.1.1.1.1.6.1 291
  46 # .1.3.6.1.4.1.10876.2.1.1.1.1.6.2 291
  47 # ...
  48 # .1.3.6.1.4.1.10876.2.1.1.1.1.11.1 RPM
  49 # .1.3.6.1.4.1.10876.2.1.1.1.1.11.2 RPM
  50 # ...
  51 # .1.3.6.1.4.1.10876.2.1.1.1.1.11.6 mV
  52 # .1.3.6.1.4.1.10876.2.1.1.1.1.11.7 mV
  53 # .1.3.6.1.4.1.10876.2.1.1.1.1.12.1 0
  54 # .1.3.6.1.4.1.10876.2.1.1.1.1.12.2 0
  55 # ...
  56 # .1.3.6.1.4.1.10876.2.1.1.1.1.12.6 0
  57 # .1.3.6.1.4.1.10876.2.1.1.1.1.12.7 0
  58 # .1.3.6.1.4.1.10876.2.2 0
  59 # .1.3.6.1.4.1.10876.2.3 No problem.
  60
  61 #.
  62 #   .--Health--------------------------------------------------------------.
  63 #   |                    _   _            _ _   _                          |
  64 #   |                   | | | | ___  __ _| | |_| |__                       |
  65 #   |                   | |_| |/ _ \/ _` | | __| '_ \                      |
  66 #   |                   |  _  |  __/ (_| | | |_| | | |                     |
  67 #   |                   |_| |_|\___|\__,_|_|\__|_| |_|                     |
  68 #   |                                                                      |
  69 #   '----------------------------------------------------------------------'
  70
  71
  72 def inventory_supermicro_health(info):
  73     if info[1]:
  74         return [(None, None)]
  75
  76
  77 def check_supermicro_health(_no_item, _no_params, info):
  78     return int(info[1][0][0]), info[1][0][1]
  79
  80
  81 check_info['supermicro'] = {
  82     'check_function': check_supermicro_health,
  83     'inventory_function': inventory_supermicro_health,
  84     'service_description': "Overall Hardware Health",
  85     'snmp_scan_function': lambda oid: oid(".1.3.6.1.2.1.1.2.0") == ".1.3.6.1.4.1.311.1.1.3.1.2",
  86     'snmp_info': [
  87         (
  88             ".1.3.6.1.4.1.10876.2.1.1.1.1",
  89             [
  90                 2,  # smHealthMonitorName
  91                 3,  # smHealthMonitorType
  92                 4,  # smHealthMonitorReading
  93                 5,  # smHealthMonitorHighLimit
  94                 6,  # smHealthMonitorLowLimit
  95                 11,  # smHealthMonitorReadingUnit
  96                 12,  # smHealthMonitorStatus
  97             ]),
  98         (
  99             ".1.3.6.1.4.1.10876.2",
 100             [
 101                 2,  # smHealthAllinoneStatus
 102                 3,  # smHealthAllinoneMsg
 103             ]),
 104         (
 105             ".1.3.6.1.4.1.10876.100.1.4.1",
 106             [
 107                 1,  # diskSerialNumber
 108                 2,  # diskName
 109                 4
 110             ]),  # diskSmartStatus
 111     ]
 112 }
 113
 114 #.
 115 #   .--Sensors-------------------------------------------------------------.
 116 #   |                 ____                                                 |
 117 #   |                / ___|  ___ _ __  ___  ___  _ __ ___                  |
 118 #   |                \___ \ / _ \ '_ \/ __|/ _ \| '__/ __|                 |
 119 #   |                 ___) |  __/ | | \__ \ (_) | |  \__ \                 |
 120 #   |                |____/ \___|_| |_|___/\___/|_|  |___/                 |
 121 #   |                                                                      |
 122 #   '----------------------------------------------------------------------'
 123
 124
 125 def inventory_supermicro_sensors(info):
 126     for name, _sensor_type, _reading, _high, _low, _unit, _status in info[0]:
 127         yield name, None
 128
 129
 130 def check_supermicro_sensors(item, _no_params, info):
 131     class Type(object):
 132         Fan, Voltage, Temperature, Status = ('0', '1', '2', '3')
 133
 134     def worst_status(*args):
 135         order = [0, 1, 3, 2]
 136         return sorted(args, key=lambda x: order[x], reverse=True)[0]
 137
 138     def expect_order(*args):
 139         return max(
 140             [abs(x[0] - x[1][0]) for x in enumerate(sorted(enumerate(args), key=lambda x: x[1]))])
 141
 142     for name, sensor_type, reading, high, low, unit, dev_status in info[0]:
 143         if name == item:
 144             reading = float(reading)
 145             dev_status = int(dev_status)
 146
 147             crit_upper = warn_upper = None
 148             status_high = status_low = 0
 149             if high:
 150                 crit_upper = float(high)
 151                 warn_upper = crit_upper * 0.95
 152                 status_high = expect_order(reading, warn_upper, crit_upper)
 153             if low:
 154                 crit_lower = float(low)
 155                 warn_lower = crit_lower * 1.05
 156                 status_low = expect_order(crit_lower, warn_lower, reading)
 157
 158             perfvar = None
 159
 160             # normalize values depending on sensor type
 161             if sensor_type == Type.Temperature:
 162                 unit = u"°%s" % unit
 163                 perfvar = "temp"
 164             elif sensor_type == Type.Voltage:
 165                 if unit == "mV":
 166                     reading, warn_upper, crit_upper =\
 167                         [x / 1000.0 for x in (reading, warn_upper, crit_upper)]
 168                     unit = "V"
 169                 perfvar = "voltage"
 170             elif sensor_type == Type.Status:
 171                 reading = "State %d" % int(reading)
 172                 unit = ""
 173
 174             perfdata = []
 175
 176             if perfvar:
 177                 if crit_upper is not None:
 178                     perfdata = [(perfvar, reading, warn_upper, crit_upper)]
 179                 else:
 180                     perfdata = [(perfvar, reading)]
 181
 182             return (worst_status(status_high, status_low, dev_status), "%s%s" % (reading, unit),
 183                     perfdata)
 184
 185
 186 check_info['supermicro.sensors'] = {
 187     'check_function': check_supermicro_sensors,
 188     'inventory_function': inventory_supermicro_sensors,
 189     'service_description': "Sensor %s",
 190     'has_perfdata': True,
 191 }
 192
 193 #.
 194 #   .--SMART---------------------------------------------------------------.
 195 #   |                   ____  __  __    _    ____ _____                    |
 196 #   |                  / ___||  \/  |  / \  |  _ \_   _|                   |
 197 #   |                  \___ \| |\/| | / _ \ | |_) || |                     |
 198 #   |                   ___) | |  | |/ ___ \|  _ < | |                     |
 199 #   |                  |____/|_|  |_/_/   \_\_| \_\|_|                     |
 200 #   |                                                                      |
 201 #   '----------------------------------------------------------------------'
 202
 203
 204 def format_item_supermicro_smart(name):
 205     return name.replace(r"\\\\.\\", "")
 206
 207
 208 def inventory_supermicro_smart(info):
 209     for _serial, name, _status in info[2]:
 210         yield format_item_supermicro_smart(name), None
 211
 212
 213 def check_supermicro_smart(item, _no_params, info):
 214     # note (only status 0 (OK) and 2 (Crit) are documented.
 215     # status 3 appears to indicate "unknown" as observed by a user.
 216     # It's likely - but not verified - that status 1 would indicate a non-
 217     # critical problem if it's used at all)
 218     status_map = {"0": "Healthy", "1": "Warning", "2": "Critical", "3": "Unknown"}
 219     for serial, name, status in info[2]:
 220         if format_item_supermicro_smart(name) == item:
 221             return int(status), "(S/N %s) %s" % (serial, status_map[status])
 222
 223
 224 check_info['supermicro.smart'] = {
 225     'check_function': check_supermicro_smart,
 226     'inventory_function': inventory_supermicro_smart,
 227     'service_description': "SMART Health %s",
 228     'has_perfdata': True,
 229 }