checks/nvidia

   1 #!/usr/bin/python
   2 # -*- encoding: utf-8; py-indent-offset: 4 -*-
   3 # +------------------------------------------------------------------+
   4 # |             ____ _               _        __  __ _  __           |
   5 # |            / ___| |__   ___  ___| | __   |  \/  | |/ /           |
   6 # |           | |   | '_ \ / _ \/ __| |/ /   | |\/| | ' /            |
   7 # |           | |___| | | |  __/ (__|   <    | |  | | . \            |
   8 # |            \____|_| |_|\___|\___|_|\_\___|_|  |_|_|\_\           |
   9 # |                                                                  |
  10 # | Copyright Mathias Kettner 2014             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 factory_settings["nvidia_temp_default_levels"] = {"levels": (60, 65)}
  28
  29 factory_settings["nvidia_temp_core_default_levels"] = {"levels": (90, 95)}
  30
  31
  32 def format_nvidia_name(identifier):
  33     identifier = identifier.replace("Temp", "")
  34     if identifier == "GPUCore":
  35         return "GPU NVIDIA"
  36
  37     # afaik temperature sensors can be GPU or Board, maybe memory
  38     return "System NVIDIA %s" % identifier
  39
  40
  41 def inventory_nvidia_temp(core, info):
  42     for line in info:
  43         line_san = line[0].strip(":")
  44         if line_san.lower().endswith("temp"):
  45             if core == (line_san == "GPUCoreTemp"):
  46                 yield format_nvidia_name(line_san), {}
  47
  48
  49 def check_nvidia_temp(item, params, info):
  50     for line in info:
  51         if format_nvidia_name(line[0].strip(":")) == item \
  52            or item == line[0].strip(":"): # compatibility code for "old discovered" services
  53             return check_temperature(int(line[1]), params, "nvidia_%s" % item)
  54
  55
  56 check_info["nvidia.temp"] = {
  57     'check_function': check_nvidia_temp,
  58     'inventory_function': lambda info: inventory_nvidia_temp(False, info),
  59     'service_description': 'Temperature %s',
  60     'has_perfdata': True,
  61     'includes': ['temperature.include'],
  62     'group': 'temperature',
  63     'default_levels_variable': 'nvidia_temp_default_levels'
  64 }
  65
  66 check_info["nvidia.temp_core"] = {
  67     'check_function': check_nvidia_temp,
  68     'inventory_function': lambda info: inventory_nvidia_temp(True, info),
  69     'service_description': 'Temperature %s',
  70     'has_perfdata': True,
  71     'includes': ['temperature.include'],
  72     'group': 'temperature',
  73     'default_levels_variable': 'nvidia_temp_core_default_levels'
  74 }
  75
  76
  77 def inventory_nvidia_errors(info):
  78     for line in info:
  79         if line[0] == 'GPUErrors:':
  80             return [(None, None)]
  81
  82
  83 def check_nvidia_errors(_no_item, _no_params, info):
  84     for line in info:
  85         if line[0] == "GPUErrors:":
  86             errors = int(line[1])
  87             if errors == 0:
  88                 return (0, "No GPU errors")
  89             return (2, "%d GPU errors" % errors)
  90     return (3, "incomplete output from agent")
  91
  92
  93 check_info["nvidia.errors"] = {
  94     'check_function': check_nvidia_errors,
  95     'inventory_function': inventory_nvidia_errors,
  96     'service_description': 'NVIDIA GPU Errors',
  97     'group': 'hw_errors',
  98 }