1 |
#!/usr/bin/python |
---|
2 |
# |
---|
3 |
# Copyright (C) 2011 Andreas Thienemann <andreas@bawue.net> |
---|
4 |
# |
---|
5 |
# This program is free software: you can redistribute it and/or modify |
---|
6 |
# it under the terms of the GNU General Public License as published by |
---|
7 |
# the Free Software Foundation, either version 3 of the License, or |
---|
8 |
# (at your option) any later version. |
---|
9 |
# |
---|
10 |
# This program is distributed in the hope that it will be useful, |
---|
11 |
# but WITHOUT ANY WARRANTY; without even the implied warranty of |
---|
12 |
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
---|
13 |
# GNU General Public License for more details. |
---|
14 |
# |
---|
15 |
# You should have received a copy of the GNU General Public License |
---|
16 |
# along with this program. If not, see <http://www.gnu.org/licenses/>. |
---|
17 |
# |
---|
18 |
|
---|
19 |
""" |
---|
20 |
=head1 NAME |
---|
21 |
|
---|
22 |
freeipmi_ - Munin plugin to retreive temperature and fan speed measurements |
---|
23 |
from a local machine via IPMI. |
---|
24 |
|
---|
25 |
=head1 APPLICABLE SYSTEMS |
---|
26 |
|
---|
27 |
All machines with an IPMI capable baseboard management controller. |
---|
28 |
|
---|
29 |
=head1 CONFIGURATION |
---|
30 |
|
---|
31 |
On most supported systems this plugin works nearly out of the box as long as |
---|
32 |
both Python and the freeipmi binaries in a semi-recent version are installed. |
---|
33 |
|
---|
34 |
If the machine works out of the box can be tested by calling bmc-info. |
---|
35 |
If there's text output, a bmc card was detected. If there's an entry for |
---|
36 |
"Sensor Device" visible in the "Additional Device Support" entry you're good. |
---|
37 |
|
---|
38 |
If you get a "ipmi_cmd_get_device_id: driver timeout" message you have most |
---|
39 |
likely no bmc to query. |
---|
40 |
|
---|
41 |
In certain cases however bmc-info will just seem to hang for quite some time. |
---|
42 |
In this case, autodetection does not work because the smbios table has |
---|
43 |
incorrect information. One system know to experience this problem is the |
---|
44 |
HP Proliant Microserver. |
---|
45 |
|
---|
46 |
Adding env.freeipmi_args "--no-probing --driver-type=KCS --driver-address=0xca2 --register-spacing=1" |
---|
47 |
to the munin plugin configuration will make the plugin work. |
---|
48 |
|
---|
49 |
Basic configuration for every system is that the plugin needs to be called as root. |
---|
50 |
|
---|
51 |
Add the following to your /etc/munin/plugin-conf.d/freeipmi: |
---|
52 |
|
---|
53 |
[freeipmi_*] |
---|
54 |
user root |
---|
55 |
|
---|
56 |
=head1 INTERPRETATION |
---|
57 |
|
---|
58 |
The plugin shows the temperature in Celsius or the fanspeed in rotations per minute. |
---|
59 |
|
---|
60 |
=head1 MAGIC MARKERS |
---|
61 |
|
---|
62 |
#%# family=contrib |
---|
63 |
#%# capabilities=autoconf suggest |
---|
64 |
|
---|
65 |
=head1 VERSION |
---|
66 |
|
---|
67 |
0.0.1 |
---|
68 |
|
---|
69 |
=head1 BUGS |
---|
70 |
|
---|
71 |
Only local support for now. Remote could be hacked in via freeipmi_args for now. |
---|
72 |
|
---|
73 |
=head1 AUTHOR |
---|
74 |
|
---|
75 |
Andreas Thienemann <andreas@bawue.net> |
---|
76 |
|
---|
77 |
=head1 LICENSE |
---|
78 |
|
---|
79 |
GPLv3+ |
---|
80 |
|
---|
81 |
=cut |
---|
82 |
""" |
---|
83 |
|
---|
84 |
import subprocess |
---|
85 |
import sys |
---|
86 |
import os |
---|
87 |
import re |
---|
88 |
import pprint |
---|
89 |
|
---|
90 |
# Parse some environment variables |
---|
91 |
if os.getenv("freeipmi_args") is not None: |
---|
92 |
freeipmi_args = " %s" % (os.getenv("freeipmi_args")) |
---|
93 |
else: |
---|
94 |
freeipmi_args = "" |
---|
95 |
|
---|
96 |
# We are a wildcard plugin, figure out whether we are called for temp or fan |
---|
97 |
if sys.argv[0].split("_")[1] == "temp": |
---|
98 |
mode = "Temperature" |
---|
99 |
elif sys.argv[0].split("_")[1] == "fan": |
---|
100 |
mode = "Fan" |
---|
101 |
else: |
---|
102 |
mode = None |
---|
103 |
|
---|
104 |
def whereis(prog): |
---|
105 |
"""Check if prog can be found in the path and if yes, return the full pathname""" |
---|
106 |
prog = os.path.basename(prog) |
---|
107 |
for dir in os.getenv("PATH").split(":"): |
---|
108 |
for root, dirs, files in os.walk(dir): |
---|
109 |
if prog in files: |
---|
110 |
return os.path.join(dir, prog) |
---|
111 |
return None |
---|
112 |
|
---|
113 |
def normalize_sensor(name): |
---|
114 |
name = name.lower().replace("-","M").replace("+","P") |
---|
115 |
name = re.sub("[^a-z0-9A-Z]","_", name) |
---|
116 |
return name |
---|
117 |
|
---|
118 |
# Code sniplet from Philipp Keller |
---|
119 |
# http://code.pui.ch/2007/02/19/set-timeout-for-a-shell-command-in-python/ |
---|
120 |
def timeout_command(command, timeout): |
---|
121 |
"""call shell-command and either return its output or kill it |
---|
122 |
if it doesn't normally exit within timeout seconds and return None""" |
---|
123 |
import subprocess, datetime, os, time, signal |
---|
124 |
start = datetime.datetime.now() |
---|
125 |
process = subprocess.Popen(command.split(), stdout=subprocess.PIPE, stderr=subprocess.PIPE) |
---|
126 |
while process.poll() is None: |
---|
127 |
time.sleep(0.1) |
---|
128 |
now = datetime.datetime.now() |
---|
129 |
if (now - start).seconds> timeout: |
---|
130 |
os.kill(process.pid, signal.SIGKILL) |
---|
131 |
os.waitpid(-1, os.WNOHANG) |
---|
132 |
return None |
---|
133 |
return process.stdout.read() |
---|
134 |
|
---|
135 |
def bmc_detect(): |
---|
136 |
"""Check whether there's a baseboard management controller we can query.""" |
---|
137 |
if whereis("bmc-info") is None: |
---|
138 |
print "no (bmc-info not found in path. Please install FreeIPMI.)" |
---|
139 |
sys.exit(0) |
---|
140 |
else: |
---|
141 |
out = timeout_command("bmc-info%s" % (freeipmi_args), 2) |
---|
142 |
if out is not None and "[Sensor Device]" in out: |
---|
143 |
print "yes" |
---|
144 |
sys.exit(0) |
---|
145 |
else: |
---|
146 |
print "no (no supported bmc found)" |
---|
147 |
sys.exit(0) |
---|
148 |
|
---|
149 |
def read_sensors(): |
---|
150 |
"""Return all sensor data as a dict""" |
---|
151 |
out = timeout_command("ipmi-sensors --verbose%s" % (freeipmi_args), 2) |
---|
152 |
sensors = dict() |
---|
153 |
sensor = dict() |
---|
154 |
sensor_id = None |
---|
155 |
for line in out.split("\n"): |
---|
156 |
if ":" in line: |
---|
157 |
k,v = line.split(": ") |
---|
158 |
if k == "Record ID": |
---|
159 |
sensor = dict() |
---|
160 |
sensor_id = int(v) |
---|
161 |
sensor[k] = v |
---|
162 |
else: |
---|
163 |
sensor[k] = v |
---|
164 |
else: |
---|
165 |
sensors[sensor_id] = sensor |
---|
166 |
return sensors |
---|
167 |
|
---|
168 |
def print_config(): |
---|
169 |
"""Return configuration arguments for munin""" |
---|
170 |
print "graph_title FreeIPMI Sensors: %s" % (mode) |
---|
171 |
if mode == "Fan": |
---|
172 |
print "graph_vlabel RPM" |
---|
173 |
print "graph_info This graph shows the RPMs of the fans as reported by IPMI" |
---|
174 |
elif mode == "Temperature": |
---|
175 |
print "graph_vlabel Degrees C" |
---|
176 |
print "graph_info This graph shows the temperatures as reported by IPMI" |
---|
177 |
print "graph_category sensors" |
---|
178 |
sensors = read_sensors() |
---|
179 |
|
---|
180 |
for id in sorted(sensors): |
---|
181 |
if sensors[id]["Group Name"] == mode: |
---|
182 |
label = normalize_sensor(sensors[id]["Sensor Name"]) |
---|
183 |
for n in ["Normal Max.", "Normal Min.", "Sensor Reading", "Lower Critical Threshold", "Upper Critical Threshold", "Lower Non-Critical Threshold", "Upper Non-Critical Threshold"]: |
---|
184 |
sensors[id][n] = sensors[id][n].replace("NA","") |
---|
185 |
sensors[id][n] = sensors[id][n].split('.')[0] |
---|
186 |
|
---|
187 |
print "%s.label %s" % (label, label) |
---|
188 |
print "%s.warning %s:%s" % (label, sensors[id]["Lower Non-Critical Threshold"], sensors[id]["Upper Non-Critical Threshold"]) |
---|
189 |
print "%s.critical %s:%s" % (label, sensors[id]["Lower Critical Threshold"], sensors[id]["Upper Critical Threshold"]) |
---|
190 |
print "%s.graph_args --base 1000 -l 0" % (label) |
---|
191 |
print "%s.graph_scale no" % (label) |
---|
192 |
# pprint.pprint(sensors[id]) |
---|
193 |
sys.exit(0) |
---|
194 |
|
---|
195 |
def fetch(): |
---|
196 |
sensors = read_sensors() |
---|
197 |
|
---|
198 |
for id in sorted(sensors): |
---|
199 |
if sensors[id]["Group Name"] == mode: |
---|
200 |
label = normalize_sensor(sensors[id]["Sensor Name"]) |
---|
201 |
print "%s.value %s" % (label, sensors[id]["Sensor Reading"].split(".")[0]) |
---|
202 |
sys.exit(0) |
---|
203 |
|
---|
204 |
|
---|
205 |
if "config" in sys.argv[1:]: |
---|
206 |
print_config() |
---|
207 |
|
---|
208 |
elif "autoconf" in sys.argv[1:]: |
---|
209 |
bmc_detect() |
---|
210 |
|
---|
211 |
elif "suggest" in sys.argv[1:]: |
---|
212 |
sensors = read_sensors() |
---|
213 |
fan, temperature = [0, 0] |
---|
214 |
for id in sensors: |
---|
215 |
if sensors[id]["Group Name"] == "Fan": |
---|
216 |
fan += 1 |
---|
217 |
elif sensors[id]["Group Name"] == "Temperature": |
---|
218 |
temperature += 1 |
---|
219 |
if fan > 0: |
---|
220 |
print "fan" |
---|
221 |
if temperature > 0: |
---|
222 |
print "temp" |
---|
223 |
sys.exit(0) |
---|
224 |
|
---|
225 |
else: |
---|
226 |
fetch() |
---|