checks/oracle_recovery_status

   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 # In cooperation with Thorsten Bruhns from OPITZ Consulting
  28
  29 # <<<oracle_recovery_state:sep(124)>>>
  30 # TUX2|tux2|PRIMARY|MOUNTED|1|1405456155|ONLINE||NO|2719061
  31 # TUX2|tux2|PRIMARY|MOUNTED|2|1405456155|ONLINE||NO|2719061
  32 # new format with backupmode
  33 # <<<oracle_recovery_status:sep(124)>>>
  34 # TUX2|tux2|PRIMARY|READ WRITE|1|1419771465|317|ONLINE|NO|YES|8149107|NOT ACTIVE|489
  35 # TUX2|tux2|PRIMARY|READ WRITE|2|1419771465|317|ONLINE|NO|YES|8149107|NOT ACTIVE|489
  36
  37 # Databases seem to also report lines with some data missing:
  38 # PV|PV|PRIMARY|READ WRITE|397|1433251398|7297|ONLINE|NO|YES|10740614283
  39 # PV|PV|PRIMARY|READ WRITE|398|1433251398|7297|ONLINE|NO|YES|10740614283
  40 # PV|PV|PRIMARY|READ WRITE|399|||ONLINE|||0
  41 # PV|PV|PRIMARY|READ WRITE|400|||ONLINE|||0
  42 # PV|PV|PRIMARY|READ WRITE|401|||ONLINE|||0
  43
  44
  45 def inventory_oracle_recovery_status(info):
  46     return [(line[0], {}) for line in info]
  47
  48
  49 def check_oracle_recovery_status(item, params, info):
  50     state = 0
  51     offlinecount = 0
  52     filemissingcount = 0
  53     oldest_checkpoint_age = None
  54
  55     oldest_backup_age = -1
  56     backup_count = 0
  57
  58     perfdata = []
  59
  60     itemfound = False
  61     for line in info:
  62         if line[0] == item:
  63             itemfound = True
  64
  65             if len(line) == 11:
  66                 db_name, db_unique_name, database_role, _open_mode, _filenr, \
  67                 _checkpoint_time, checkpoint_age, datafilestatus, _recovery, _fuzzy, _checkpoint_change = line
  68
  69                 backup_state = 'unknown'
  70
  71             elif len(line) == 13:
  72                 db_name, db_unique_name, database_role, _open_mode, _filenr, \
  73                 _checkpoint_time, checkpoint_age, datafilestatus, _recovery, \
  74                 _fuzzy, _checkpoint_change, backup_state, backup_age = line
  75
  76             if params.get("levels"):
  77                 warn, crit = params["levels"]
  78
  79             if backup_state == 'ACTIVE':
  80                 backup_count += 1
  81                 oldest_backup_age = max(int(backup_age), oldest_backup_age)
  82
  83             if datafilestatus == 'ONLINE':
  84                 if backup_state == 'FILE MISSING':
  85                     filemissingcount += 1
  86                 elif checkpoint_age:
  87                     checkpoint_age = int(checkpoint_age)
  88
  89                     if oldest_checkpoint_age is None:
  90                         oldest_checkpoint_age = checkpoint_age
  91                     else:
  92                         oldest_checkpoint_age = max(oldest_checkpoint_age, checkpoint_age)
  93
  94             else:
  95                 offlinecount += 1
  96
  97     if itemfound is True:
  98         infotext = "%s database" % (database_role.lower())
  99
 100         if oldest_checkpoint_age is None:
 101             infotext += ", no online datafiles found(!!)"
 102             state = 2
 103
 104         elif oldest_checkpoint_age <= -1:
 105             # we found a negative time for last checkpoint
 106             infotext += ", oldest checkpoint is in the future  %s(!), check the time on the server" \
 107                         % get_age_human_readable(int(oldest_checkpoint_age)*-1)
 108             state = max(state, 1)
 109
 110         else:
 111             infotext += ", oldest Checkpoint %s ago" \
 112                        % (get_age_human_readable(int(oldest_checkpoint_age)))
 113
 114         if (((database_role == 'PRIMARY' and db_name == '_MGMTDB' and db_unique_name == '_mgmtdb') \
 115              or not params.get("levels")
 116             )
 117             or db_name[db_name.rfind('.')+1:] == 'PDB$SEED'
 118            ):
 119
 120             # We ignore the state of the check when no parameters are known
 121             # _mgmtdb is new internal instance from 12.1.0.2 on Grid-Infrastructure
 122             # ignore PDB$SEED because this PDB is always in READ ONLY mode
 123             perfdata.append(('checkpoint_age', oldest_checkpoint_age))
 124         else:
 125             if database_role == 'PRIMARY':
 126                 # checkpoint age should not higher on primary as well
 127                 # There is no CRIT for older checkoint age as this is mostly not a
 128                 # serios issue.
 129                 # otherwise the standby will produca a warning or crit as well
 130                 if oldest_checkpoint_age >= warn:
 131                     infotext += '(!)'
 132                     state = max(1, state)
 133
 134                 perfdata.append(('checkpoint_age', oldest_checkpoint_age, warn))
 135             else:
 136                 perfdata.append(('checkpoint_age', oldest_checkpoint_age, warn, crit))
 137
 138                 # check the checkpoint age on a non primary database!
 139                 if oldest_checkpoint_age >= crit:
 140                     infotext += '(!!)'
 141                     state = 2
 142                 elif oldest_checkpoint_age >= warn:
 143                     infotext += '(!)'
 144                     state = max(1, state)
 145
 146             infotext += ' (warn/crit at %s/%s )' % (get_age_human_readable(warn),
 147                                                     get_age_human_readable(crit))
 148
 149         if offlinecount > 0:
 150             infotext += " %i datafiles offline(!!)" \
 151                         % (offlinecount)
 152             state = 2
 153
 154         if filemissingcount > 0:
 155             infotext += " %i missing datafiles(!!)" \
 156                         % (filemissingcount)
 157             state = 2
 158
 159         if oldest_backup_age > 0:
 160             infotext += " %i datafiles in backup mode oldest is %s" % ( \
 161                          backup_count, get_age_human_readable(oldest_backup_age))
 162
 163             if params.get("backup_age"):
 164
 165                 warn, crit = params["backup_age"]
 166                 infotext += " (warn/crit at %s/%s)" % (get_age_human_readable(warn),
 167                                                        get_age_human_readable(crit))
 168                 perfdata.append(('backup_age', oldest_backup_age, warn, crit))
 169
 170                 if oldest_backup_age >= crit:
 171                     infotext += '(!!)'
 172                     state = 2
 173                 elif oldest_backup_age >= warn:
 174                     infotext += '(!)'
 175                     state = max(1, state)
 176             else:
 177                 perfdata.append(('backup_age', oldest_backup_age))
 178         else:
 179
 180             # create a 'dummy' performance data with 0
 181             # => The age from plugin is only valid when a datafile is in backup mode!
 182             perfdata.append(('backup_age', 0))
 183
 184         return state, infotext, perfdata
 185
 186     # In case of missing information we assume that the login into
 187     # the database has failed and we simply skip this check. It won't
 188     # switch to UNKNOWN, but will get stale.
 189     raise MKCounterWrapped("Login into database failed")
 190
 191
 192 check_info['oracle_recovery_status'] = {
 193     "check_function": check_oracle_recovery_status,
 194     "inventory_function": inventory_oracle_recovery_status,
 195     "service_description": "ORA %s Recovery Status",
 196     "has_perfdata": True,
 197     "group": "oracle_recovery_status",
 198 }