checks/kernel

   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 #   .--kernel--Counters----------------------------------------------------.
  28 #   |                ____                  _                               |
  29 #   |               / ___|___  _   _ _ __ | |_ ___ _ __ ___                |
  30 #   |              | |   / _ \| | | | '_ \| __/ _ \ '__/ __|               |
  31 #   |              | |__| (_) | |_| | | | | ||  __/ |  \__ \               |
  32 #   |               \____\___/ \__,_|_| |_|\__\___|_|  |___/               |
  33 #   |                                                                      |
  34 #   +----------------------------------------------------------------------+
  35 #   |  Check page faults, context switches and process creations           |
  36 #   '----------------------------------------------------------------------'
  37
  38 # Inventory creates three checks per default:
  39 inventory_kernel_counters = ["pgmajfault", "ctxt", "processes"]
  40 kernel_default_levels = None
  41
  42 kernel_counter_names = {
  43     "ctxt": "Context Switches",
  44     "processes": "Process Creations",
  45     "pgmajfault": "Major Page Faults",
  46 }
  47
  48
  49 def inventory_kernel(info):
  50     inventory = []
  51     for counter in inventory_kernel_counters:
  52         hits = [line[0] for line in info[1:] if line[0] == counter]
  53         if len(hits) == 1:
  54             countername = kernel_counter_names.get(counter, counter)
  55             inventory.append((countername, "kernel_default_levels"))
  56     return inventory
  57
  58
  59 # item is one of the keys in /proc/stat or /proc/vmstat
  60 def check_kernel(item, params, info):
  61     if not info:
  62         return
  63
  64     this_time = int(info[0][0])
  65     hits = [(line[0], line[1])
  66             for line in info[1:]
  67             if line[0] == item or kernel_counter_names.get(line[0], line[0]) == item]
  68     if len(hits) == 0:
  69         return (3, "item '%s' not found in agent output" % item)
  70     elif len(hits) > 1:
  71         return (3, "item '%s' not unique (found %d times)" % (item, len(hits)))
  72
  73     counter = hits[0][0]
  74     this_val = int(hits[0][1])
  75     per_sec = get_rate(None, this_time, this_val)
  76
  77     return check_levels(per_sec, counter, params, unit="/s", infoname=kernel_counter_names[counter])
  78
  79
  80 check_info["kernel"] = {
  81     'check_function': check_kernel,
  82     'inventory_function': inventory_kernel,
  83     'service_description': 'Kernel %s',
  84     'has_perfdata': True,
  85     'group': 'vm_counter',
  86 }
  87
  88 #   .--kernel.util--CPU Utilization----------------------------------------.
  89 #   |             _   _ _   _ _ _          _   _                           |
  90 #   |            | | | | |_(_) (_)______ _| |_(_) ___  _ __                |
  91 #   |            | | | | __| | | |_  / _` | __| |/ _ \| '_ \               |
  92 #   |            | |_| | |_| | | |/ / (_| | |_| | (_) | | | |              |
  93 #   |             \___/ \__|_|_|_/___\__,_|\__|_|\___/|_| |_|              |
  94 #   |                                                                      |
  95 #   +----------------------------------------------------------------------+
  96 #   |  Check system/user/io-wait                                           |
  97 #   '----------------------------------------------------------------------'
  98
  99
 100 def inventory_cpu_utilization(info):
 101     for x in info:
 102         if len(x) > 0 and x[0] == 'cpu':
 103             return [(None, {})]
 104
 105
 106 # Columns of cpu usage /proc/stat:
 107 # - cpuX: number of CPU or only 'cpu' for aggregation
 108 # - user: normal processes executing in user mode
 109 # - nice: niced processes executing in user mode
 110 # - system: processes executing in kernel mode
 111 # - idle: twiddling thumbs
 112 # - iowait: waiting for I/O to complete
 113 # - irq: servicing interrupts
 114 # - softirq: servicing softirqs
 115 # - steal: Stolen time, which is the time spent in other operating systems
 116 #          when running in a virtualized environment (since Linux 2.6.11)
 117 # - guest: Time spent running a virtual CPU for guest operating systems (since Linux 2.6.24)
 118 # - guest_nice: Time spent running a niced guest (since Linux 2.6.33)
 119
 120
 121 def kernel_check_cpu_utilization(item, params, info):
 122     params = transform_cpu_iowait(params)
 123
 124     # Look for entry matching "cpu" (this is the combined load of all cores)
 125     total = [cpu_info(line) for line in info if line[0] == "cpu"]
 126
 127     if len(total) != 1:
 128         return 3, "More than one line with CPU info found. This check is not cluster-enabled."
 129
 130     cores = [cpu_info(line) for line in info if line[0].startswith("cpu") and len(line[0]) > 3]
 131     # total contains now the following columns:
 132     # 'cpu' user nice system idle wait hw-int sw-int (steal ...)
 133     # convert number to int
 134     return check_cpu_util_unix(total[0], params, cores)
 135
 136
 137 check_info["kernel.util"] = {
 138     'check_function': kernel_check_cpu_utilization,
 139     'inventory_function': inventory_cpu_utilization,
 140     'service_description': 'CPU utilization',
 141     'has_perfdata': True,
 142     'group': 'cpu_iowait',
 143     'includes': ['cpu_util.include', 'transforms.include'],
 144 }