1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
|
# -*- coding: utf-8 -*-
# Description: nvidia-smi netdata python.d module
# Original Author: Steven Noonan (tycho)
# Author: Ilya Mashchenko (ilyam8)
import subprocess
import threading
import xml.etree.ElementTree as et
from bases.collection import find_binary
from bases.FrameworkServices.SimpleService import SimpleService
disabled_by_default = True
NVIDIA_SMI = 'nvidia-smi'
BAD_VALUE = 'N/A'
EMPTY_ROW = ''
EMPTY_ROW_LIMIT = 500
POLLER_BREAK_ROW = '</nvidia_smi_log>'
PCI_BANDWIDTH = 'pci_bandwidth'
FAN_SPEED = 'fan_speed'
GPU_UTIL = 'gpu_utilization'
MEM_UTIL = 'mem_utilization'
ENCODER_UTIL = 'encoder_utilization'
MEM_USAGE = 'mem_usage'
TEMPERATURE = 'temperature'
CLOCKS = 'clocks'
POWER = 'power'
PROCESSES_MEM = 'processes_mem'
ORDER = [
PCI_BANDWIDTH,
FAN_SPEED,
GPU_UTIL,
MEM_UTIL,
ENCODER_UTIL,
MEM_USAGE,
TEMPERATURE,
CLOCKS,
POWER,
PROCESSES_MEM,
]
def gpu_charts(gpu):
fam = gpu.full_name()
charts = {
PCI_BANDWIDTH: {
'options': [None, 'PCI Express Bandwidth Utilization', 'KiB/s', fam, 'nvidia_smi.pci_bandwidth', 'area'],
'lines': [
['rx_util', 'rx', 'absolute', 1, 1],
['tx_util', 'tx', 'absolute', 1, -1],
]
},
FAN_SPEED: {
'options': [None, 'Fan Speed', 'percentage', fam, 'nvidia_smi.fan_speed', 'line'],
'lines': [
['fan_speed', 'speed'],
]
},
GPU_UTIL: {
'options': [None, 'GPU Utilization', 'percentage', fam, 'nvidia_smi.gpu_utilization', 'line'],
'lines': [
['gpu_util', 'utilization'],
]
},
MEM_UTIL: {
'options': [None, 'Memory Bandwidth Utilization', 'percentage', fam, 'nvidia_smi.mem_utilization', 'line'],
'lines': [
['memory_util', 'utilization'],
]
},
ENCODER_UTIL: {
'options': [None, 'Encoder/Decoder Utilization', 'percentage', fam, 'nvidia_smi.encoder_utilization', 'line'],
'lines': [
['encoder_util', 'encoder'],
['decoder_util', 'decoder'],
]
},
MEM_USAGE: {
'options': [None, 'Memory Usage', 'MiB', fam, 'nvidia_smi.memory_allocated', 'stacked'],
'lines': [
['fb_memory_free', 'free'],
['fb_memory_used', 'used'],
]
},
TEMPERATURE: {
'options': [None, 'Temperature', 'celsius', fam, 'nvidia_smi.temperature', 'line'],
'lines': [
['gpu_temp', 'temp'],
]
},
CLOCKS: {
'options': [None, 'Clock Frequencies', 'MHz', fam, 'nvidia_smi.clocks', 'line'],
'lines': [
['graphics_clock', 'graphics'],
['video_clock', 'video'],
['sm_clock', 'sm'],
['mem_clock', 'mem'],
]
},
POWER: {
'options': [None, 'Power Utilization', 'Watts', fam, 'nvidia_smi.power', 'line'],
'lines': [
['power_draw', 'power', 'absolute', 1, 100],
]
},
PROCESSES_MEM: {
'options': [None, 'Memory Used by Each Process', 'MiB', fam, 'nvidia_smi.processes_mem', 'stacked'],
'lines': []
},
}
idx = gpu.num
order = ['gpu{0}_{1}'.format(idx, v) for v in ORDER]
charts = dict(('gpu{0}_{1}'.format(idx, k), v) for k, v in charts.items())
for chart in charts.values():
for line in chart['lines']:
line[0] = 'gpu{0}_{1}'.format(idx, line[0])
return order, charts
class NvidiaSMI:
def __init__(self):
self.command = find_binary(NVIDIA_SMI)
self.active_proc = None
def run_once(self):
proc = subprocess.Popen([self.command, '-x', '-q'], stdout=subprocess.PIPE)
stdout, _ = proc.communicate()
return stdout
def run_loop(self, interval):
if self.active_proc:
self.kill()
proc = subprocess.Popen([self.command, '-x', '-q', '-l', str(interval)], stdout=subprocess.PIPE)
self.active_proc = proc
return proc.stdout
def kill(self):
if self.active_proc:
self.active_proc.kill()
self.active_proc = None
class NvidiaSMIPoller(threading.Thread):
def __init__(self, poll_interval):
threading.Thread.__init__(self)
self.daemon = True
self.smi = NvidiaSMI()
self.interval = poll_interval
self.lock = threading.RLock()
self.last_data = str()
self.exit = False
self.empty_rows = 0
self.rows = list()
def has_smi(self):
return bool(self.smi.command)
def run_once(self):
return self.smi.run_once()
def run(self):
out = self.smi.run_loop(self.interval)
for row in out:
if self.exit or self.empty_rows > EMPTY_ROW_LIMIT:
break
self.process_row(row)
self.smi.kill()
def process_row(self, row):
row = row.decode()
self.empty_rows += (row == EMPTY_ROW)
self.rows.append(row)
if POLLER_BREAK_ROW in row:
self.lock.acquire()
self.last_data = '\n'.join(self.rows)
self.lock.release()
self.rows = list()
self.empty_rows = 0
def is_started(self):
return self.ident is not None
def shutdown(self):
self.exit = True
def data(self):
self.lock.acquire()
data = self.last_data
self.lock.release()
return data
def handle_attr_error(method):
def on_call(*args, **kwargs):
try:
return method(*args, **kwargs)
except AttributeError:
return None
return on_call
def handle_value_error(method):
def on_call(*args, **kwargs):
try:
return method(*args, **kwargs)
except ValueError:
return None
return on_call
class GPU:
def __init__(self, num, root):
self.num = num
self.root = root
def id(self):
return self.root.get('id')
def name(self):
return self.root.find('product_name').text
def full_name(self):
return 'gpu{0} {1}'.format(self.num, self.name())
@handle_attr_error
def rx_util(self):
return self.root.find('pci').find('rx_util').text.split()[0]
@handle_attr_error
def tx_util(self):
return self.root.find('pci').find('tx_util').text.split()[0]
@handle_attr_error
def fan_speed(self):
return self.root.find('fan_speed').text.split()[0]
@handle_attr_error
def gpu_util(self):
return self.root.find('utilization').find('gpu_util').text.split()[0]
@handle_attr_error
def memory_util(self):
return self.root.find('utilization').find('memory_util').text.split()[0]
@handle_attr_error
def encoder_util(self):
return self.root.find('utilization').find('encoder_util').text.split()[0]
@handle_attr_error
def decoder_util(self):
return self.root.find('utilization').find('decoder_util').text.split()[0]
@handle_attr_error
def fb_memory_used(self):
return self.root.find('fb_memory_usage').find('used').text.split()[0]
@handle_attr_error
def fb_memory_free(self):
return self.root.find('fb_memory_usage').find('free').text.split()[0]
@handle_attr_error
def temperature(self):
return self.root.find('temperature').find('gpu_temp').text.split
|