libexec/check_shinken.py

   1 #!/usr/bin/env python
   2 #Copyright (C) 2009-2011 :
   3 #    Denis GERMAIN, dt.germain@gmail.com
   4 #    Gabes Jean, naparuba@gmail.com
   5 #    Gerhard Lausser, Gerhard.Lausser@consol.de
   6 #    Gregory Starck, g.starck@gmail.com
   7 #    Hartmut Goebel, h.goebel@goebel-consult.de
   8 #
   9 #This file is part of Shinken.
  10 #
  11 #Shinken is free software: you can redistribute it and/or modify
  12 #it under the terms of the GNU Affero General Public License as published by
  13 #the Free Software Foundation, either version 3 of the License, or
  14 #(at your option) any later version.
  15 #
  16 #Shinken is distributed in the hope that it will be useful,
  17 #but WITHOUT ANY WARRANTY; without even the implied warranty of
  18 #MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  19 #GNU Affero General Public License for more details.
  20 #
  21 #You should have received a copy of the GNU Affero General Public License
  22 #along with Shinken.  If not, see <http://www.gnu.org/licenses/>.
  23
  24 ################################################
  25 # check_shinken.py :
  26 #    This check is getting daemons state from
  27 #    a arbiter connexion.
  28 ################################################
  29
  30 import os
  31
  32 # Exit statuses recognized by Nagios and thus by Shinken
  33 OK = 0
  34 WARNING = 1
  35 CRITICAL = 2
  36 UNKNOWN = 3
  37
  38 #Name of the Pyro Object we are searching
  39 PYRO_OBJECT = 'ForArbiter'
  40 daemon_types = ['arbiter', 'broker', 'scheduler', 'poller', 'reactionner']
  41
  42 try:
  43     import shinken
  44 except ImportError:
  45     # If importing shinken fails, try to load from current directory
  46     # or parent directory to support running without installation.
  47     # Submodules will then be loaded from there, too.
  48     import imp
  49     if not hasattr(os, "getuid") or os.getuid() != 0:
  50         imp.load_module('shinken', *imp.find_module('shinken', [".", ".."]))
  51
  52
  53 from optparse import OptionParser
  54 try:
  55     import shinken.pyro_wrapper as pyro
  56     from shinken.pyro_wrapper import Pyro
  57 except ImportError, exp:
  58     print 'CRITICAL : check_shinken requires the Python Pyro and the shinken.pyro_wrapper module. Please install it. (%s)' % exp
  59     raise SystemExit, CRITICAL
  60
  61
  62 def check_deamons_numbers(result, target):
  63     total_number = len(result)
  64     alive_number = len([e for e in result.values() if e['alive']])
  65     total_spare_number = len([e for e in result.values() if e['spare']])
  66     alive_spare_number = len([e for e in result.values() if e['spare'] and e['alive']])
  67     dead_number = total_number - alive_number
  68     dead_list = ','.join([n for n in result if not result[n]['alive']])
  69     #TODO : perfdata to graph deamons would be nice (in big HA architectures)
  70     #if alive_number <= critical, then we have a big problem
  71     if alive_number <= options.critical:
  72         print "CRITICAL - only %d/%d %s(s) UP. Down elements : %s" % (alive_number, total_number, target, dead_list)
  73         raise SystemExit, CRITICAL
  74     #We are not in a case where there is no more daemons, but are there daemons down?
  75     elif dead_number >= options.warning:
  76         print "WARNING - %d/%d %s(s) DOWN :%s" % (dead_number, total_number, target, dead_list)
  77         raise SystemExit, WARNING
  78         #Everything seems fine. But that's no surprise, is it?
  79     else :
  80         print "OK - %d/%d %s(s) UP, with %d/%d spare(s) UP" % (alive_number, total_number, target, alive_spare_number, total_spare_number)
  81         raise SystemExit, OK
  82
  83 # Adding options. None are required, check_shinken will use shinken defaults
  84 #TODO : Add more control in args problem and usage than the default OptionParser one
  85 parser = OptionParser()
  86 parser.add_option('-a', '--hostname', dest='hostname', default='127.0.0.1')
  87 parser.add_option('-p', '--portnumber', dest='portnum', default=7770)
  88 parser.add_option('-s', '--ssl', dest='ssl', default=False)
  89 #TODO : Add a list of correct values for target and don't authorize anything else
  90 parser.add_option('-t', '--target', dest='target')
  91 parser.add_option('-d', '--daemonname', dest='daemon', default='')
  92 #In HA architectures, a warning should be displayed if there's one daemon down
  93 parser.add_option('-w','--warning', dest='warning', default = 1)
  94 #If no deamon is left, display a critical (but shinken will be probably dead already)
  95 parser.add_option('-c', '--critical', dest='critical', default = 0)
  96
  97 #Retrieving options
  98 options, args = parser.parse_args()
  99 #TODO : for now, helpme doesn't work as desired
 100 options.helpme = False
 101
 102 # Check for required option target
 103 if not getattr(options, 'target'):
 104     print 'CRITICAL - target is not specified; You must specify which daemons you want to check!'
 105     parser.print_help()
 106     raise SystemExit, CRITICAL
 107 elif options.target not in daemon_types:
 108     print 'CRITICAL - target %s is not a Shinken daemon!' % options.target
 109     parser.print_help()
 110     raise SystemExit, CRITICAL
 111
 112 uri = pyro.create_uri(options.hostname, options.portnum, PYRO_OBJECT , options.ssl)
 113
 114 if options.daemon:
 115     # We just want a check for a single satellite daemon
 116     # Only OK or CRITICAL here
 117     daemon_name = options.daemon
 118     try:
 119         result = Pyro.core.getProxyForURI(uri).get_satellite_status(options.target, daemon_name)
 120     except Pyro.errors.ProtocolError, exp:
 121         print "CRITICAL : the Arbiter is not reachable : (%s)." % exp
 122         raise SystemExit, CRITICAL
 123
 124     if result:
 125         if result['alive']:
 126             print 'OK - %s alive' % daemon_name
 127             raise SystemExit, OK
 128         else:
 129             print 'CRITICAL - %s down' % daemon_name
 130             raise SystemExit, CRITICAL
 131     else:
 132         print 'UNKNOWN - %s status could not be retrieved' % daemon_name
 133         raise SystemExit, UNKNOWN
 134 else:
 135     # If no daemonname is specified, we want a general overview of the "target" daemons
 136     result = {}
 137
 138     try:
 139         daemon_list = Pyro.core.getProxyForURI(uri).get_satellite_list(options.target)
 140     except Pyro.errors.ProtocolError, exp:
 141         print "CRITICAL : the Arbiter is not reachable : (%s)." % exp
 142         raise SystemExit, CRITICAL
 143
 144     for daemon_name in daemon_list:
 145         # Getting individual daemon and putting status info in the result dictionnary
 146         try:
 147             result[daemon_name] = Pyro.core.getProxyForURI(uri).get_satellite_status(options.target, daemon_name)
 148         except Pyro.errors.ProtocolError, exp:
 149             print "CRITICAL : the Arbiter is not reachable : (%s)." % exp
 150             raise SystemExit, CRITICAL
 151
 152     # Now we have all data
 153     if result:
 154         check_deamons_numbers(result, options.target)
 155     else :
 156         print 'UNKNOWN - Arbiter could not retrieve status for %s' % options.target
 157         raise SystemExit, UNKNOWN