checks/zpool_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 # Based on check from  Darin Perusich <darin@darins.net>
  28 #
  29 # /usr/sbin/zpool status -x
  30 #
  31 # Example output of a healthy spool
  32 # ---------------------------------
  33 # all pools are healthy
  34 # ---------------------------------
  35 #
  36 # Example output with no poole
  37 # ---------------------------------
  38 # no pools available
  39 # ---------------------------------
  40 #
  41 # Example output of a pool with a error
  42 # ---------------------------------
  43 # pool: snapshots
  44 #  state: ONLINE
  45 # status: One or more devices has experienced an unrecoverable error.  An
  46 #         attempt was made to correct the error.  Applications are unaffected.
  47 # action: Determine if the device needs to be replaced, and clear the errors
  48 #         using 'zpool clear' or replace the device with 'zpool replace'.
  49 #    see: http://www.sun.com/msg/ZFS-8000-9P
  50 #  scrub: none requested
  51 # config:
  52 #
  53 #         NAME        STATE     READ WRITE CKSUM
  54 #         snapshots   ONLINE       0     0     0
  55 #           raidz1    ONLINE       0     0     0
  56 #             c0d0s7  ONLINE       0     0     0
  57 #             c0d1s7  ONLINE       0     0     0
  58 #             c1d0s7  ONLINE       0     0     0
  59 #             c1d1s7  ONLINE       0     0     1
  60 #
  61 # errors: No known data errors
  62 # ---------------------------------
  63
  64
  65 def inventory_zpool_status(info):
  66     if not info or " ".join(info[0]) == "no pools available":
  67         return []
  68     return [(None, None)]
  69
  70
  71 def check_zpool_status(_no_item, _no_params, info):
  72     if " ".join(info[0]) == "all pools are healthy":
  73         return 0, "All pools are healthy"
  74     elif " ".join(info[0]) == "no pools available":
  75         return 3, "No pools available"
  76
  77     start_pool    = False
  78     multiline     = False
  79     last_pool     = None
  80     error_pools   = {}
  81     warning_pools = {}
  82     pool_messages = {}
  83     message = []
  84     state   = 0
  85
  86     for line in info:
  87         if line[0] == "pool:":
  88             last_pool = line[1]
  89             pool_messages.setdefault(last_pool, [])
  90             continue
  91
  92         if line[0] == "state:":
  93             if   line[1] == "ONLINE":
  94                  state = 0
  95             elif line[1] == "DEGRADED":
  96                  state = 1
  97                  message.append("DEGRADED State")
  98             elif line[1] == "FAULTED":
  99                  state = 2
 100                  message.append("FAULTED State")
 101             elif line[1] == "UNAVIL":
 102                  state = 2
 103                  message.append("UNAVIL State")
 104             elif line[1] == "REMOVED":
 105                  state = 2
 106                  message.append("REMOVED State")
 107             elif line[1] == "OFFLINE":
 108                  state = 0
 109             else:
 110                  message.append("Unknown State")
 111                  state = 1
 112             continue
 113
 114         if line[0] in ["status:", "action:"]:
 115             pool_messages[last_pool].append(" ".join(line[1:]))
 116             multiline = True
 117             continue
 118
 119         if line[0] in ["scrub:", "see:", "scan:", "config:"]:
 120             multiline = False
 121             continue
 122
 123         if line[0] == "NAME":
 124             multiline  = False
 125             start_pool = True
 126             continue
 127
 128         if line[0] == "errors:":
 129             multiline  = False
 130             start_pool = False
 131             msg = " ".join(line[1:])
 132             if msg != 'No known data errors':
 133                 pool_messages[last_pool].append(msg)
 134             continue
 135
 136         if line[0] in ["spares", "logs", "cache"]:
 137             start_pool = False
 138             continue
 139
 140         if start_pool is True:
 141             if line[1] != "ONLINE":
 142                 error_pools[line[0]] = tuple(line[1:])
 143                 continue
 144
 145             if saveint(line[4]) != 0:
 146                 warning_pools[line[0]] = tuple(line[1:])
 147                 continue
 148
 149         if multiline:
 150             pool_messages[last_pool].append(" ".join(line))
 151             continue
 152
 153     for pool, msg in pool_messages.iteritems():
 154         state = 1
 155         message.append("%s: %s" % (pool, " ".join(msg)))
 156
 157     for pool, msg in warning_pools.iteritems():
 158         state = 1
 159         message.append("%s CKSUM: %d" % (pool, saveint(msg[3])))
 160
 161     for pool, msg in error_pools.iteritems():
 162         state = 2
 163         message.append("%s state: %s" % (pool, msg[0]))
 164
 165     if len(message) == 0:
 166         message = ['No critical errors']
 167
 168     return state, ", ".join(message)
 169
 170
 171 check_info["zpool_status"] = {
 172     'check_function'      : check_zpool_status,
 173     'inventory_function'  : inventory_zpool_status,
 174     'service_description' : 'zpool status',
 175     'group'               : 'zpool_status',
 176 }