check_ceph_usage

Initial commit
Benjamin Renard

Benjamin Renard commited on 2013-12-03 10:00:39
Showing 3 changed files, with 255 additions and 0 deletions.

.gitignore
View file @ fadfa27
... ... 
@@ -0,0 +1 @@
1 
+*~
README
View file @ fadfa27
... ... 
@@ -0,0 +1,47 @@
1 
+Nagios plugin to check Ceph cluster usage
2 
+=========================================
3 
+
4 
+Usage
5 
+-----
6 
+
7 
+    Usage: check_ceph_usage [options]
8 
+
9 
+    Options:
10 
+      -h, --help            show this help message and exit
11 
+      -d, --debug
12 
+      -b BIN, --bin=BIN     Ceph binary (default : /usr/bin/ceph)
13 
+      --conf=CONF           Ceph configuration file
14 
+      -m MON, --mon=MON     Ceph monitor address[:port]
15 
+      -i ID, --id=ID        Ceph client id
16 
+      -k KEYRING, --keyring=KEYRING
17 
+                            Ceph client keyring file
18 
+      -w WARNDATA, --warning-data=WARNDATA
19 
+                            Warning data threshold (default : 70%)
20 
+      -c CRITDATA, --critical-data=CRITDATA
21 
+                            Critical data threshold (default : 85%)
22 
+      -W WARNALLOC, --warning-allocated=WARNALLOC
23 
+                            Warning allocated threshold (default : 80%)
24 
+      -C CRITALLOC, --critical-allocated=CRITALLOC
25 
+                            Critical allocated threshold (default : 90%)
26 
+
27 
+Copyright
28 
+---------
29 
+
30 
+Copyright (c) 2013 Benjamin Renard <brenard@zionetrix.net>
31 
+
32 
+License
33 
+-------
34 
+
35 
+This program is free software; you can redistribute it and/or
36 
+modify it under the terms of the GNU General Public License version 2
37 
+as published by the Free Software Foundation.
38 
+
39 
+This program is distributed in the hope that it will be useful,
40 
+but WITHOUT ANY WARRANTY; without even the implied warranty of
41 
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
42 
+GNU General Public License for more details.
43 
+
44 
+You should have received a copy of the GNU General Public License
45 
+along with this program; if not, write to the Free Software
46 
+Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA  02111-1307, USA.
47 
+
check_ceph_usage
View file @ fadfa27
... ... 
@@ -0,0 +1,207 @@
1 
+#!/usr/bin/python
2 
+#
3 
+# Nagios plugin to check Ceph cluster usage
4 
+#
5 
+#     Usage: check_ceph_usage [options]
6 
+#
7 
+#     Options:
8 
+#       -h, --help            show this help message and exit
9 
+#       -d, --debug
10 
+#       -b BIN, --bin=BIN     Ceph binary (default : /usr/bin/ceph)
11 
+#       --conf=CONF           Ceph configuration file
12 
+#       -m MON, --mon=MON     Ceph monitor address[:port]
13 
+#       -i ID, --id=ID        Ceph client id
14 
+#       -k KEYRING, --keyring=KEYRING
15 
+#                             Ceph client keyring file
16 
+#       -w WARNDATA, --warning-data=WARNDATA
17 
+#                             Warning data threshold (default : 70%)
18 
+#       -c CRITDATA, --critical-data=CRITDATA
19 
+#                             Critical data threshold (default : 85%)
20 
+#       -W WARNALLOC, --warning-allocated=WARNALLOC
21 
+#                             Warning allocated threshold (default : 80%)
22 
+#       -C CRITALLOC, --critical-allocated=CRITALLOC
23 
+#                             Critical allocated threshold (default : 90%)
24 
+#
25 
+# Copyright (c) 2013 Benjamin Renard <brenard@zionetrix.net>
26 
+#
27 
+# This program is free software; you can redistribute it and/or
28 
+# modify it under the terms of the GNU General Public License version 2
29 
+# as published by the Free Software Foundation.
30 
+#
31 
+# This program is distributed in the hope that it will be useful,
32 
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
33 
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
34 
+# GNU General Public License for more details.
35 
+#
36 
+# You should have received a copy of the GNU General Public License
37 
+# along with this program; if not, write to the Free Software
38 
+# Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA  02111-1307, USA.
39 
+#
40 
+
41 
+import sys,os,json,subprocess
42 
+from optparse import OptionParser
43 
+
44 
+# default ceph values
45 
+CEPH_COMMAND = '/usr/bin/ceph'
46 
+WARN_DATA = 70
47 
+CRIT_DATA = 85
48 
+WARN_ALLOC = 80
49 
+CRIT_ALLOC = 90
50 
+
51 
+# nagios exit code
52 
+STATUS_OK = 0
53 
+STATUS_WARNING = 1
54 
+STATUS_ERROR = 2
55 
+STATUS_UNKNOWN = 3
56 
+
57 
+parser = OptionParser()
58 
+parser.add_option('-d',
59 
+                  '--debug',
60 
+                  action="store_true",
61 
+                  dest="debug",
62 
+                  default=False)
63 
+
64 
+parser.add_option('-b',
65 
+                  '--bin',
66 
+                  action="store",
67 
+                  dest="bin",
68 
+                  help="Ceph binary (default : %s)" % CEPH_COMMAND,
69 
+                  type='string',
70 
+                  default=CEPH_COMMAND)
71 
+
72 
+parser.add_option('--conf',
73 
+                  action="store",
74 
+                  dest="conf",
75 
+                  help="Ceph configuration file",
76 
+                  type='string',
77 
+                  default=None)
78 
+
79 
+parser.add_option('-m',
80 
+                  '--mon',
81 
+                  action="store",
82 
+                  dest="mon",
83 
+                  help="Ceph monitor address[:port]",
84 
+                  type='string',
85 
+                  default=None)
86 
+
87 
+parser.add_option('-i',
88 
+                  '--id',
89 
+                  action="store",
90 
+                  dest="id",
91 
+                  help="Ceph client id",
92 
+                  type='string',
93 
+                  default=None)
94 
+
95 
+parser.add_option('-k',
96 
+                  '--keyring',
97 
+                  action="store",
98 
+                  dest="keyring",
99 
+                  help="Ceph client keyring file",
100 
+                  type='string',
101 
+                  default=None)
102 
+
103 
+parser.add_option('-w',
104 
+                  '--warning-data',
105 
+                  action="store",
106 
+                  dest="warndata",
107 
+                  help="Warning data threshold (default : %s%%)" % WARN_DATA,
108 
+                  type='int',
109 
+                  default=WARN_DATA)
110 
+
111 
+parser.add_option('-c',
112 
+                  '--critical-data',
113 
+                  action="store",
114 
+                  dest="critdata",
115 
+                  help="Critical data threshold (default : %s%%)" % CRIT_DATA,
116 
+                  type='int',
117 
+                  default=CRIT_DATA)
118 
+
119 
+parser.add_option('-W',
120 
+                  '--warning-allocated',
121 
+                  action="store",
122 
+                  dest="warnalloc",
123 
+                  help="Warning allocated threshold (default : %s%%)" % WARN_ALLOC,
124 
+                  type='int',
125 
+                  default=WARN_ALLOC)
126 
+
127 
+parser.add_option('-C',
128 
+                  '--critical-allocated',
129 
+                  action="store",
130 
+                  dest="critalloc",
131 
+                  help="Critical allocated threshold (default : %s%%)" % CRIT_ALLOC,
132 
+                  type='int',
133 
+                  default=CRIT_ALLOC)
134 
+
135 
+(options, args) = parser.parse_args()
136 
+
137 
+ # validate args
138 
+if not os.path.exists(options.bin):
139 
+    print "ERROR: ceph executable '%s' doesn't exist" % options.bin
140 
+    sys.exit(STATUS_UNKNOWN)
141 
+
142 
+if options.conf and not os.path.exists(options.conf):
143 
+    print "ERROR: ceph conf file '%s' doesn't exist" % options.conf
144 
+    sys.exit(STATUS_UNKNOWN)
145 
+
146 
+if options.keyring and not os.path.exists(options.keyring):
147 
+    print "ERROR: keyring file '%s' doesn't exist" % options.keyring
148 
+    sys.exit(STATUS_UNKNOWN)
149 
+
150 
+# build command
151 
+ceph_cmd = [options.bin]
152 
+if options.mon:
153 
+    ceph_cmd.append('-m')
154 
+    ceph_cmd.append(options.mon)
155 
+if options.conf:
156 
+    ceph_cmd.append('-c')
157 
+    ceph_cmd.append(options.conf)
158 
+if options.id:
159 
+    ceph_cmd.append('--id')
160 
+    ceph_cmd.append(options.id)
161 
+if options.keyring:
162 
+    ceph_cmd.append('--keyring')
163 
+    ceph_cmd.append(options.keyring)
164 
+ceph_cmd.append('status')
165 
+ceph_cmd.append('--format=json')
166 
+
167 
+# exec command
168 
+p = subprocess.Popen(ceph_cmd,stdout=subprocess.PIPE,stderr=subprocess.PIPE)
169 
+output, err = p.communicate()
170 
+
171 
+if output:
172 
+	data=json.loads(output)
173 
+	if 'pgmap' not in data:
174 
+		print "UNKNOWN : fail to read pgmap status"
175 
+		sys.exit(STATUS_UNKNOWN)
176 
+
177 
+	if options.debug:
178 
+		print "data : %s" % data['pgmap']['data_bytes']
179 
+		print "allocated : %s" % data['pgmap']['bytes_used']
180 
+		print "total : %s" % data['pgmap']['bytes_total']
181 
+
182 
+	PER_DATA=round(int(data['pgmap']['data_bytes'])*100/int(data['pgmap']['bytes_total']),1)
183 
+	DATA_WARN_T=int(int(data['pgmap']['bytes_total'])*options.warndata/100)
184 
+	DATA_CRIT_T=int(int(data['pgmap']['bytes_total'])*options.critdata/100)
185 
+	PER_ALLOC=round(int(data['pgmap']['bytes_used'])*100/int(data['pgmap']['bytes_total']),1)
186 
+	ALLOC_WARN_T=int(int(data['pgmap']['bytes_total'])*options.warnalloc/100)
187 
+	ALLOC_CRIT_T=int(int(data['pgmap']['bytes_total'])*options.critalloc/100)
188 
+
189 
+
190 
+	if options.debug:
191 
+		print "%% data : %s" % PER_DATA
192 
+		print "%% allocated : %s" % PER_ALLOC
193 
+
194 
+	STATUS=STATUS_OK
195 
+	STATUS_TXT="OK"
196 
+	if PER_DATA > options.critdata or PER_ALLOC > options.critalloc:
197 
+		STATUS=STATUS_CRITICAL
198 
+		STATUS_TXT="CRITICAL"
199 
+	elif PER_DATA > options.warndata or PER_ALLOC > options.warnalloc:
200 
+		STATUS=STATUS_WARNING
201 
+		STATUS_TXT="WARNING"
202 
+
203 
+	print "%s : %s%% allocated / %s%% really used|allocated=%sB;%s;%s;0;%s,used=%sB;%s;%s;0;%s" % (STATUS_TXT,PER_ALLOC,PER_DATA,data['pgmap']['bytes_used'],ALLOC_WARN_T,ALLOC_CRIT_T,data['pgmap']['bytes_total'],data['pgmap']['data_bytes'],DATA_WARN_T,DATA_CRIT_T,data['pgmap']['bytes_total'])
204 
+	sys.exit(STATUS)
205 
+else:
206 
+	print "UNKNOWN : fail to execute ceph status command"
207 
+	sys.exit(STATUS_UNKNOWN)
0 208