fix label issues with prometheus api+clean stop of metric export
[osm/vim-emu.git] / src / emuvim / dcemulator / monitoring.py
1 """
2 Copyright (c) 2015 SONATA-NFV
3 ALL RIGHTS RESERVED.
4
5 Licensed under the Apache License, Version 2.0 (the "License");
6 you may not use this file except in compliance with the License.
7 You may obtain a copy of the License at
8
9 http://www.apache.org/licenses/LICENSE-2.0
10
11 Unless required by applicable law or agreed to in writing, software
12 distributed under the License is distributed on an "AS IS" BASIS,
13 WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
14 See the License for the specific language governing permissions and
15 limitations under the License.
16
17 Neither the name of the SONATA-NFV [, ANY ADDITIONAL AFFILIATION]
18 nor the names of its contributors may be used to endorse or promote
19 products derived from this software without specific prior written
20 permission.
21
22 This work has been performed in the framework of the SONATA project,
23 funded by the European Commission under Grant number 671517 through
24 the Horizon 2020 and 5G-PPP programmes. The authors would like to
25 acknowledge the contributions of their colleagues of the SONATA
26 partner consortium (www.sonata-nfv.eu).
27 """
28
29 import logging
30 import sys
31 from mininet.node import OVSSwitch
32 import ast
33 import time
34 from prometheus_client import start_http_server, Summary, Histogram, Gauge, Counter, REGISTRY, CollectorRegistry, \
35 pushadd_to_gateway, push_to_gateway, delete_from_gateway
36 import threading
37 from subprocess import Popen
38 import os
39 import docker
40
41 logging.basicConfig(level=logging.INFO)
42
43 """
44 class to read openflow stats from the Ryu controller of the DCNetwork
45 """
46
47 PUSHGATEWAY_PORT = 9091
48 # we cannot use port 8080 because ryu-ofrest api is already using that one
49 CADVISOR_PORT = 8081
50
51 COOKIE_MASK = 0xffffffff
52
53 class DCNetworkMonitor():
54 def __init__(self, net):
55 self.net = net
56 self.dockercli = docker.from_env()
57
58 # pushgateway address
59 self.pushgateway = 'localhost:{0}'.format(PUSHGATEWAY_PORT)
60
61 # supported Prometheus metrics
62 self.registry = CollectorRegistry()
63 self.prom_tx_packet_count = Gauge('sonemu_tx_count_packets', 'Total number of packets sent',
64 ['vnf_name', 'vnf_interface', 'flow_id'], registry=self.registry)
65 self.prom_rx_packet_count = Gauge('sonemu_rx_count_packets', 'Total number of packets received',
66 ['vnf_name', 'vnf_interface', 'flow_id'], registry=self.registry)
67 self.prom_tx_byte_count = Gauge('sonemu_tx_count_bytes', 'Total number of bytes sent',
68 ['vnf_name', 'vnf_interface', 'flow_id'], registry=self.registry)
69 self.prom_rx_byte_count = Gauge('sonemu_rx_count_bytes', 'Total number of bytes received',
70 ['vnf_name', 'vnf_interface', 'flow_id'], registry=self.registry)
71
72 self.prom_metrics={'tx_packets':self.prom_tx_packet_count, 'rx_packets':self.prom_rx_packet_count,
73 'tx_bytes':self.prom_tx_byte_count,'rx_bytes':self.prom_rx_byte_count}
74
75 # list of installed metrics to monitor
76 # each entry can contain this data
77 '''
78 {
79 switch_dpid = 0
80 vnf_name = None
81 vnf_interface = None
82 previous_measurement = 0
83 previous_monitor_time = 0
84 metric_key = None
85 mon_port = None
86 }
87 '''
88 self.monitor_lock = threading.Lock()
89 self.monitor_flow_lock = threading.Lock()
90 self.network_metrics = []
91 self.flow_metrics = []
92
93 # start monitoring thread
94 self.start_monitoring = True
95 self.monitor_thread = threading.Thread(target=self.get_network_metrics)
96 self.monitor_thread.start()
97
98 self.monitor_flow_thread = threading.Thread(target=self.get_flow_metrics)
99 self.monitor_flow_thread.start()
100
101 # helper tools
102 # cAdvisor, Prometheus pushgateway are started as external container, to gather monitoring metric in son-emu
103 self.pushgateway_process = self.start_PushGateway()
104 self.cadvisor_process = self.start_cAdvisor()
105
106
107 # first set some parameters, before measurement can start
108 def setup_flow(self, vnf_name, vnf_interface=None, metric='tx_packets', cookie=0):
109
110 flow_metric = {}
111
112 # check if port is specified (vnf:port)
113 if vnf_interface is None:
114 # take first interface by default
115 connected_sw = self.net.DCNetwork_graph.neighbors(vnf_name)[0]
116 link_dict = self.net.DCNetwork_graph[vnf_name][connected_sw]
117 vnf_interface = link_dict[0]['src_port_id']
118
119 flow_metric['vnf_name'] = vnf_name
120 flow_metric['vnf_interface'] = vnf_interface
121
122 vnf_switch = None
123 for connected_sw in self.net.DCNetwork_graph.neighbors(vnf_name):
124 link_dict = self.net.DCNetwork_graph[vnf_name][connected_sw]
125 for link in link_dict:
126 if link_dict[link]['src_port_id'] == vnf_interface:
127 # found the right link and connected switch
128 vnf_switch = connected_sw
129 flow_metric['mon_port'] = link_dict[link]['dst_port_nr']
130 break
131
132 if not vnf_switch:
133 logging.exception("vnf switch of {0}:{1} not found!".format(vnf_name, vnf_interface))
134 return "vnf switch of {0}:{1} not found!".format(vnf_name, vnf_interface)
135
136 try:
137 # default port direction to monitor
138 if metric is None:
139 metric = 'tx_packets'
140
141 next_node = self.net.getNodeByName(vnf_switch)
142
143 if not isinstance(next_node, OVSSwitch):
144 logging.info("vnf: {0} is not connected to switch".format(vnf_name))
145 return
146
147 flow_metric['previous_measurement'] = 0
148 flow_metric['previous_monitor_time'] = 0
149
150 flow_metric['switch_dpid'] = int(str(next_node.dpid), 16)
151 flow_metric['metric_key'] = metric
152 flow_metric['cookie'] = cookie
153
154 self.monitor_flow_lock.acquire()
155 self.flow_metrics.append(flow_metric)
156 self.monitor_flow_lock.release()
157
158 logging.info('Started monitoring flow:{3} {2} on {0}:{1}'.format(vnf_name, vnf_interface, metric, cookie))
159 return 'Started monitoring flow:{3} {2} on {0}:{1}'.format(vnf_name, vnf_interface, metric, cookie)
160
161 except Exception as ex:
162 logging.exception("setup_metric error.")
163 return ex.message
164
165 def stop_flow(self, vnf_name, vnf_interface=None, metric=None, cookie=0,):
166
167 # check if port is specified (vnf:port)
168 if vnf_interface is None and metric is not None:
169 # take first interface by default
170 connected_sw = self.net.DCNetwork_graph.neighbors(vnf_name)[0]
171 link_dict = self.net.DCNetwork_graph[vnf_name][connected_sw]
172 vnf_interface = link_dict[0]['src_port_id']
173
174 for flow_dict in self.flow_metrics:
175 if flow_dict['vnf_name'] == vnf_name and flow_dict['vnf_interface'] == vnf_interface \
176 and flow_dict['metric_key'] == metric and flow_dict['cookie'] == cookie:
177
178 self.monitor_flow_lock.acquire()
179
180 self.flow_metrics.remove(flow_dict)
181
182 # set metric to NaN
183 self.prom_metrics[flow_dict['metric_key']]. \
184 labels(vnf_name=vnf_name, vnf_interface=vnf_interface, flow_id=cookie). \
185 set(float('nan'))
186
187 delete_from_gateway(self.pushgateway, job='sonemu-SDNcontroller')
188
189 self.monitor_flow_lock.release()
190
191 logging.info('Stopped monitoring flow {3}: {2} on {0}:{1}'.format(vnf_name, vnf_interface, metric, cookie))
192 return 'Stopped monitoring flow {3}: {2} on {0}:{1}'.format(vnf_name, vnf_interface, metric, cookie)
193
194 return 'Error stopping monitoring flow: {0} on {1}:{2}'.format(metric, vnf_name, vnf_interface)
195
196
197 # first set some parameters, before measurement can start
198 def setup_metric(self, vnf_name, vnf_interface=None, metric='tx_packets'):
199
200 network_metric = {}
201
202 # check if port is specified (vnf:port)
203 if vnf_interface is None:
204 # take first interface by default
205 connected_sw = self.net.DCNetwork_graph.neighbors(vnf_name)[0]
206 link_dict = self.net.DCNetwork_graph[vnf_name][connected_sw]
207 vnf_interface = link_dict[0]['src_port_id']
208
209 network_metric['vnf_name'] = vnf_name
210 network_metric['vnf_interface'] = vnf_interface
211
212 for connected_sw in self.net.DCNetwork_graph.neighbors(vnf_name):
213 link_dict = self.net.DCNetwork_graph[vnf_name][connected_sw]
214 for link in link_dict:
215 if link_dict[link]['src_port_id'] == vnf_interface:
216 # found the right link and connected switch
217 network_metric['mon_port'] = link_dict[link]['dst_port_nr']
218 break
219
220 if 'mon_port' not in network_metric:
221 logging.exception("vnf interface {0}:{1} not found!".format(vnf_name,vnf_interface))
222 return "vnf interface {0}:{1} not found!".format(vnf_name,vnf_interface)
223
224 try:
225 # default port direction to monitor
226 if metric is None:
227 metric = 'tx_packets'
228
229 vnf_switch = self.net.DCNetwork_graph.neighbors(str(vnf_name))
230
231 if len(vnf_switch) > 1:
232 logging.info("vnf: {0} has multiple ports".format(vnf_name))
233 return
234 elif len(vnf_switch) == 0:
235 logging.info("vnf: {0} is not connected".format(vnf_name))
236 return
237 else:
238 vnf_switch = vnf_switch[0]
239 next_node = self.net.getNodeByName(vnf_switch)
240
241 if not isinstance(next_node, OVSSwitch):
242 logging.info("vnf: {0} is not connected to switch".format(vnf_name))
243 return
244
245 network_metric['previous_measurement'] = 0
246 network_metric['previous_monitor_time'] = 0
247
248
249 network_metric['switch_dpid'] = int(str(next_node.dpid), 16)
250 network_metric['metric_key'] = metric
251
252 self.monitor_lock.acquire()
253 self.network_metrics.append(network_metric)
254 self.monitor_lock.release()
255
256
257 logging.info('Started monitoring: {2} on {0}:{1}'.format(vnf_name, vnf_interface, metric))
258 return 'Started monitoring: {2} on {0}:{1}'.format(vnf_name, vnf_interface, metric)
259
260 except Exception as ex:
261 logging.exception("setup_metric error.")
262 return ex.message
263
264 def stop_metric(self, vnf_name, vnf_interface=None, metric=None):
265
266 # check if port is specified (vnf:port)
267 if vnf_interface is None and metric is not None:
268 # take first interface by default
269 connected_sw = self.net.DCNetwork_graph.neighbors(vnf_name)[0]
270 link_dict = self.net.DCNetwork_graph[vnf_name][connected_sw]
271 vnf_interface = link_dict[0]['src_port_id']
272
273 for metric_dict in self.network_metrics:
274 if metric_dict['vnf_name'] == vnf_name and metric_dict['vnf_interface'] == vnf_interface \
275 and metric_dict['metric_key'] == metric:
276
277 self.monitor_lock.acquire()
278
279 self.network_metrics.remove(metric_dict)
280
281 # set values to NaN, prometheus api currently does not support removal of metrics
282 #self.prom_metrics[metric_dict['metric_key']].labels(vnf_name, vnf_interface).set(float('nan'))
283 self.prom_metrics[metric_dict['metric_key']]. \
284 labels(vnf_name=vnf_name, vnf_interface=vnf_interface, flow_id=None). \
285 set(float('nan'))
286
287 # this removes the complete metric, all labels...
288 # 1 single monitor job for all metrics of the SDN controller
289 # we can only remove from the pushgateway grouping keys(labels) which we have defined for the add_to_pushgateway
290 # we can not specify labels from the metrics to be removed
291 # if we need to remove the metrics seperatelty, we need to give them a separate grouping key, and probably a diffferent registry also
292 delete_from_gateway(self.pushgateway, job='sonemu-SDNcontroller')
293
294 self.monitor_lock.release()
295
296 logging.info('Stopped monitoring: {2} on {0}:{1}'.format(vnf_name, vnf_interface, metric))
297 return 'Stopped monitoring: {2} on {0}:{1}'.format(vnf_name, vnf_interface, metric)
298
299 # delete everything from this vnf
300 elif metric_dict['vnf_name'] == vnf_name and vnf_interface is None and metric is None:
301 self.monitor_lock.acquire()
302 self.network_metrics.remove(metric_dict)
303 for collector in self.registry._collectors:
304 collector_dict = collector._metrics.copy()
305 for name, interface, id in collector_dict:
306 if name == vnf_name:
307 logging.info('3 name:{0} labels:{1} metrics:{2}'.format(collector._name, collector._labelnames,
308 collector._metrics))
309 collector.remove(name, interface, 'None')
310
311 delete_from_gateway(self.pushgateway, job='sonemu-SDNcontroller')
312 self.monitor_lock.release()
313 logging.info('Stopped monitoring vnf: {0}'.format(vnf_name))
314 return 'Stopped monitoring: {0}'.format(vnf_name)
315
316 return 'Error stopping monitoring metric: {0} on {1}:{2}'.format(metric, vnf_name, vnf_interface)
317
318
319 # get all metrics defined in the list and export it to Prometheus
320 def get_flow_metrics(self):
321 while self.start_monitoring:
322
323 self.monitor_flow_lock.acquire()
324
325 for flow_dict in self.flow_metrics:
326 data = {}
327
328 data['cookie'] = flow_dict['cookie']
329 data['cookie_mask'] = COOKIE_MASK
330
331 if 'tx' in flow_dict['metric_key']:
332 data['match'] = {'in_port':flow_dict['mon_port']}
333 elif 'rx' in flow_dict['metric_key']:
334 data['out_port'] = flow_dict['mon_port']
335
336
337 # query Ryu
338 ret = self.net.ryu_REST('stats/flow', dpid=flow_dict['switch_dpid'], data=data)
339 if isinstance(ret, dict):
340 flow_stat_dict = ret
341 elif isinstance(ret, basestring):
342 flow_stat_dict = ast.literal_eval(ret.rstrip())
343 else:
344 flow_stat_dict = None
345
346 logging.debug('received flow stat:{0} '.format(flow_stat_dict))
347
348 self.set_flow_metric(flow_dict, flow_stat_dict)
349
350
351 try:
352 if len(self.flow_metrics) > 0:
353 pushadd_to_gateway(self.pushgateway, job='sonemu-SDNcontroller', registry=self.registry)
354 except Exception, e:
355 logging.warning("Pushgateway not reachable: {0} {1}".format(Exception, e))
356
357 self.monitor_flow_lock.release()
358 time.sleep(1)
359
360 def get_network_metrics(self):
361 while self.start_monitoring:
362
363 self.monitor_lock.acquire()
364
365 # group metrics by dpid to optimize the rest api calls
366 dpid_list = [metric_dict['switch_dpid'] for metric_dict in self.network_metrics]
367 dpid_set = set(dpid_list)
368
369 for dpid in dpid_set:
370
371 # query Ryu
372 ret = self.net.ryu_REST('stats/port', dpid=dpid)
373 if isinstance(ret, dict):
374 port_stat_dict = ret
375 elif isinstance(ret, basestring):
376 port_stat_dict = ast.literal_eval(ret.rstrip())
377 else:
378 port_stat_dict = None
379
380 metric_list = [metric_dict for metric_dict in self.network_metrics
381 if int(metric_dict['switch_dpid'])==int(dpid)]
382
383 for metric_dict in metric_list:
384 self.set_network_metric(metric_dict, port_stat_dict)
385
386 try:
387 if len(self.network_metrics) > 0:
388 pushadd_to_gateway(self.pushgateway, job='sonemu-SDNcontroller', registry=self.registry)
389 except Exception, e:
390 logging.warning("Pushgateway not reachable: {0} {1}".format(Exception, e))
391
392 self.monitor_lock.release()
393 time.sleep(1)
394
395 # add metric to the list to export to Prometheus, parse the Ryu port-stats reply
396 def set_network_metric(self, metric_dict, port_stat_dict):
397 # vnf tx is the datacenter switch rx and vice-versa
398 metric_key = self.switch_tx_rx(metric_dict['metric_key'])
399 switch_dpid = metric_dict['switch_dpid']
400 vnf_name = metric_dict['vnf_name']
401 vnf_interface = metric_dict['vnf_interface']
402 previous_measurement = metric_dict['previous_measurement']
403 previous_monitor_time = metric_dict['previous_monitor_time']
404 mon_port = metric_dict['mon_port']
405
406 for port_stat in port_stat_dict[str(switch_dpid)]:
407 if int(port_stat['port_no']) == int(mon_port):
408 port_uptime = port_stat['duration_sec'] + port_stat['duration_nsec'] * 10 ** (-9)
409 this_measurement = int(port_stat[metric_key])
410
411 # set prometheus metric
412 self.prom_metrics[metric_dict['metric_key']].\
413 labels(vnf_name=vnf_name, vnf_interface=vnf_interface, flow_id=None).\
414 set(this_measurement)
415
416 # also the rate is calculated here, but not used for now
417 # (rate can be easily queried from prometheus also)
418 if previous_monitor_time <= 0 or previous_monitor_time >= port_uptime:
419 metric_dict['previous_measurement'] = int(port_stat[metric_key])
420 metric_dict['previous_monitor_time'] = port_uptime
421 # do first measurement
422 #time.sleep(1)
423 #self.monitor_lock.release()
424 # rate cannot be calculated yet (need a first measurement)
425 metric_rate = None
426
427 else:
428 time_delta = (port_uptime - metric_dict['previous_monitor_time'])
429 metric_rate = (this_measurement - metric_dict['previous_measurement']) / float(time_delta)
430
431 metric_dict['previous_measurement'] = this_measurement
432 metric_dict['previous_monitor_time'] = port_uptime
433 return
434
435 logging.exception('metric {0} not found on {1}:{2}'.format(metric_key, vnf_name, vnf_interface))
436 logging.exception('monport:{0}, dpid:{1}'.format(mon_port, switch_dpid))
437 logging.exception('port dict:{0}'.format(port_stat_dict))
438 return 'metric {0} not found on {1}:{2}'.format(metric_key, vnf_name, vnf_interface)
439
440 def set_flow_metric(self, metric_dict, flow_stat_dict):
441 # vnf tx is the datacenter switch rx and vice-versa
442 metric_key = metric_dict['metric_key']
443 switch_dpid = metric_dict['switch_dpid']
444 vnf_name = metric_dict['vnf_name']
445 vnf_interface = metric_dict['vnf_interface']
446 previous_measurement = metric_dict['previous_measurement']
447 previous_monitor_time = metric_dict['previous_monitor_time']
448 cookie = metric_dict['cookie']
449
450 counter = 0
451 for flow_stat in flow_stat_dict[str(switch_dpid)]:
452 if 'bytes' in metric_key:
453 counter += flow_stat['byte_count']
454 elif 'packet' in metric_key:
455 counter += flow_stat['packet_count']
456
457 # flow_uptime disabled for now (can give error)
458 #flow_stat = flow_stat_dict[str(switch_dpid)][0]
459 #flow_uptime = flow_stat['duration_sec'] + flow_stat['duration_nsec'] * 10 ** (-9)
460
461 self.prom_metrics[metric_dict['metric_key']]. \
462 labels(vnf_name=vnf_name, vnf_interface=vnf_interface, flow_id=cookie). \
463 set(counter)
464
465 def start_Prometheus(self, port=9090):
466 # prometheus.yml configuration file is located in the same directory as this file
467 cmd = ["docker",
468 "run",
469 "--rm",
470 "-p", "{0}:9090".format(port),
471 "-v", "{0}/prometheus.yml:/etc/prometheus/prometheus.yml".format(os.path.dirname(os.path.abspath(__file__))),
472 "-v", "{0}/profile.rules:/etc/prometheus/profile.rules".format(os.path.dirname(os.path.abspath(__file__))),
473 "--name", "prometheus",
474 "prom/prometheus"
475 ]
476 logging.info('Start Prometheus container {0}'.format(cmd))
477 return Popen(cmd)
478
479 def start_PushGateway(self, port=PUSHGATEWAY_PORT):
480 cmd = ["docker",
481 "run",
482 "-d",
483 "-p", "{0}:9091".format(port),
484 "--name", "pushgateway",
485 "--label", 'com.containernet=""',
486 "prom/pushgateway"
487 ]
488
489 logging.info('Start Prometheus Push Gateway container {0}'.format(cmd))
490 return Popen(cmd)
491
492 def start_cAdvisor(self, port=CADVISOR_PORT):
493 cmd = ["docker",
494 "run",
495 "--rm",
496 "--volume=/:/rootfs:ro",
497 "--volume=/var/run:/var/run:rw",
498 "--volume=/sys:/sys:ro",
499 "--volume=/var/lib/docker/:/var/lib/docker:ro",
500 "--publish={0}:8080".format(port),
501 "--name=cadvisor",
502 "--label",'com.containernet=""',
503 "google/cadvisor:latest"
504 ]
505 logging.info('Start cAdvisor container {0}'.format(cmd))
506 return Popen(cmd)
507
508 def stop(self):
509 # stop the monitoring thread
510 self.start_monitoring = False
511 self.monitor_thread.join()
512 self.monitor_flow_thread.join()
513
514 # these containers are used for monitoring but are started now outside of son-emu
515
516 if self.pushgateway_process is not None:
517 logging.info('stopping pushgateway container')
518 self._stop_container('pushgateway')
519
520 if self.cadvisor_process is not None:
521 logging.info('stopping cadvisor container')
522 self._stop_container('cadvisor')
523
524 def switch_tx_rx(self,metric=''):
525 # when monitoring vnfs, the tx of the datacenter switch is actually the rx of the vnf
526 # so we need to change the metric name to be consistent with the vnf rx or tx
527 if 'tx' in metric:
528 metric = metric.replace('tx','rx')
529 elif 'rx' in metric:
530 metric = metric.replace('rx','tx')
531
532 return metric
533
534 def _stop_container(self, name):
535
536 container = self.dockercli.containers.get(name)
537 container.remove(force=True)
538
539