src/emuvim/dcemulator/monitoring.py

   1 __author__ = 'Administrator'
   2
   3 import urllib2
   4 import logging
   5 from mininet.node import  OVSSwitch
   6 import ast
   7 import time
   8 from prometheus_client import start_http_server, Summary, Histogram, Gauge, Counter, REGISTRY, CollectorRegistry, \
   9     pushadd_to_gateway, push_to_gateway, delete_from_gateway
  10 import threading
  11 from subprocess import Popen, PIPE
  12 import os
  13
  14 import paramiko
  15 import gevent
  16
  17 logging.basicConfig(level=logging.INFO)
  18
  19 """
  20 class to read openflow stats from the Ryu controller of the DCNetwork
  21 """
  22
  23 class DCNetworkMonitor():
  24     def __init__(self, net):
  25         self.net = net
  26
  27         prometheus_ip = '0.0.0.0'
  28         prometheus_port = '9090'
  29         self.prometheus_REST_api = 'http://{0}:{1}'.format(prometheus_ip, prometheus_port)
  30
  31
  32
  33         # helper variables to calculate the metrics
  34         self.pushgateway = 'localhost:9091'
  35         # Start up the server to expose the metrics to Prometheus.
  36         #start_http_server(8000)
  37         # supported Prometheus metrics
  38         self.registry = CollectorRegistry()
  39         self.prom_tx_packet_count = Gauge('sonemu_tx_count_packets', 'Total number of packets sent',
  40                                           ['vnf_name', 'vnf_interface', 'flow_id'], registry=self.registry)
  41         self.prom_rx_packet_count = Gauge('sonemu_rx_count_packets', 'Total number of packets received',
  42                                           ['vnf_name', 'vnf_interface', 'flow_id'], registry=self.registry)
  43         self.prom_tx_byte_count = Gauge('sonemu_tx_count_bytes', 'Total number of bytes sent',
  44                                         ['vnf_name', 'vnf_interface', 'flow_id'], registry=self.registry)
  45         self.prom_rx_byte_count = Gauge('sonemu_rx_count_bytes', 'Total number of bytes received',
  46                                         ['vnf_name', 'vnf_interface', 'flow_id'], registry=self.registry)
  47
  48         self.prom_metrics={'tx_packets':self.prom_tx_packet_count, 'rx_packets':self.prom_rx_packet_count,
  49                            'tx_bytes':self.prom_tx_byte_count,'rx_bytes':self.prom_rx_byte_count}
  50
  51         # list of installed metrics to monitor
  52         # each entry can contain this data
  53         '''
  54         {
  55         switch_dpid = 0
  56         vnf_name = None
  57         vnf_interface = None
  58         previous_measurement = 0
  59         previous_monitor_time = 0
  60         metric_key = None
  61         mon_port = None
  62         }
  63         '''
  64         self.monitor_lock = threading.Lock()
  65         self.monitor_flow_lock = threading.Lock()
  66         self.network_metrics = []
  67         self.flow_metrics = []
  68
  69         # start monitoring thread
  70         self.start_monitoring = True
  71         self.monitor_thread = threading.Thread(target=self.get_network_metrics)
  72         self.monitor_thread.start()
  73
  74         self.monitor_flow_thread = threading.Thread(target=self.get_flow_metrics)
  75         self.monitor_flow_thread.start()
  76
  77         # helper tools
  78         self.pushgateway_process = self.start_PushGateway()
  79         self.prometheus_process = self.start_Prometheus()
  80         self.cadvisor_process = self.start_cadvisor()
  81
  82     # first set some parameters, before measurement can start
  83     def setup_flow(self, vnf_name, vnf_interface=None, metric='tx_packets', cookie=0):
  84
  85         flow_metric = {}
  86
  87         # check if port is specified (vnf:port)
  88         if vnf_interface is None:
  89             # take first interface by default
  90             connected_sw = self.net.DCNetwork_graph.neighbors(vnf_name)[0]
  91             link_dict = self.net.DCNetwork_graph[vnf_name][connected_sw]
  92             vnf_interface = link_dict[0]['src_port_id']
  93
  94         flow_metric['vnf_name'] = vnf_name
  95         flow_metric['vnf_interface'] = vnf_interface
  96
  97         vnf_switch = None
  98         for connected_sw in self.net.DCNetwork_graph.neighbors(vnf_name):
  99             link_dict = self.net.DCNetwork_graph[vnf_name][connected_sw]
 100             for link in link_dict:
 101                 # logging.info("{0},{1}".format(link_dict[link],vnf_interface))
 102                 if link_dict[link]['src_port_id'] == vnf_interface:
 103                     # found the right link and connected switch
 104                     # logging.info("{0},{1}".format(link_dict[link]['src_port_id'], vnf_source_interface))
 105                     vnf_switch = connected_sw
 106                     flow_metric['mon_port'] = link_dict[link]['dst_port_nr']
 107                     break
 108
 109         if not vnf_switch:
 110             logging.exception("vnf switch of {0}:{1} not found!".format(vnf_name, vnf_interface))
 111             return "vnf switch of {0}:{1} not found!".format(vnf_name, vnf_interface)
 112
 113         try:
 114             # default port direction to monitor
 115             if metric is None:
 116                 metric = 'tx_packets'
 117
 118             next_node = self.net.getNodeByName(vnf_switch)
 119
 120             if not isinstance(next_node, OVSSwitch):
 121                 logging.info("vnf: {0} is not connected to switch".format(vnf_name))
 122                 return
 123
 124             flow_metric['previous_measurement'] = 0
 125             flow_metric['previous_monitor_time'] = 0
 126
 127             flow_metric['switch_dpid'] = int(str(next_node.dpid), 16)
 128             flow_metric['metric_key'] = metric
 129             flow_metric['cookie'] = cookie
 130
 131             self.monitor_flow_lock.acquire()
 132             self.flow_metrics.append(flow_metric)
 133             self.monitor_flow_lock.release()
 134
 135             logging.info('Started monitoring flow:{3} {2} on {0}:{1}'.format(vnf_name, vnf_interface, metric, cookie))
 136             return 'Started monitoring flow:{3} {2} on {0}:{1}'.format(vnf_name, vnf_interface, metric, cookie)
 137
 138         except Exception as ex:
 139             logging.exception("setup_metric error.")
 140             return ex.message
 141
 142
 143     # first set some parameters, before measurement can start
 144     def setup_metric(self, vnf_name, vnf_interface=None, metric='tx_packets'):
 145
 146         network_metric = {}
 147
 148         # check if port is specified (vnf:port)
 149         if vnf_interface is None:
 150             # take first interface by default
 151             connected_sw = self.net.DCNetwork_graph.neighbors(vnf_name)[0]
 152             link_dict = self.net.DCNetwork_graph[vnf_name][connected_sw]
 153             vnf_interface = link_dict[0]['src_port_id']
 154
 155         network_metric['vnf_name'] = vnf_name
 156         network_metric['vnf_interface'] = vnf_interface
 157
 158         for connected_sw in self.net.DCNetwork_graph.neighbors(vnf_name):
 159             link_dict = self.net.DCNetwork_graph[vnf_name][connected_sw]
 160             for link in link_dict:
 161                 # logging.info("{0},{1}".format(link_dict[link],vnf_interface))
 162                 if link_dict[link]['src_port_id'] == vnf_interface:
 163                     # found the right link and connected switch
 164                     # logging.info("{0},{1}".format(link_dict[link]['src_port_id'], vnf_source_interface))
 165                     network_metric['mon_port'] = link_dict[link]['dst_port_nr']
 166                     break
 167
 168         if 'mon_port' not in network_metric:
 169             logging.exception("vnf interface {0}:{1} not found!".format(vnf_name,vnf_interface))
 170             return "vnf interface {0}:{1} not found!".format(vnf_name,vnf_interface)
 171
 172         try:
 173             # default port direction to monitor
 174             if metric is None:
 175                 metric = 'tx_packets'
 176
 177             vnf_switch = self.net.DCNetwork_graph.neighbors(str(vnf_name))
 178
 179             if len(vnf_switch) > 1:
 180                 logging.info("vnf: {0} has multiple ports".format(vnf_name))
 181                 return
 182             elif len(vnf_switch) == 0:
 183                 logging.info("vnf: {0} is not connected".format(vnf_name))
 184                 return
 185             else:
 186                 vnf_switch = vnf_switch[0]
 187             next_node = self.net.getNodeByName(vnf_switch)
 188
 189             if not isinstance(next_node, OVSSwitch):
 190                 logging.info("vnf: {0} is not connected to switch".format(vnf_name))
 191                 return
 192
 193             network_metric['previous_measurement'] = 0
 194             network_metric['previous_monitor_time'] = 0
 195
 196
 197             network_metric['switch_dpid'] = int(str(next_node.dpid), 16)
 198             network_metric['metric_key'] = metric
 199
 200             self.monitor_lock.acquire()
 201
 202             self.network_metrics.append(network_metric)
 203             self.monitor_lock.release()
 204
 205
 206             logging.info('Started monitoring: {2} on {0}:{1}'.format(vnf_name, vnf_interface, metric))
 207             return 'Started monitoring: {2} on {0}:{1}'.format(vnf_name, vnf_interface, metric)
 208
 209         except Exception as ex:
 210             logging.exception("setup_metric error.")
 211             return ex.message
 212
 213     def stop_metric(self, vnf_name, vnf_interface=None, metric=None):
 214
 215         for metric_dict in self.network_metrics:
 216             #logging.info('start Stopped monitoring: {2} on {0}:{1}'.format(vnf_name, vnf_interface, metric_dict))
 217             if metric_dict['vnf_name'] == vnf_name and metric_dict['vnf_interface'] == vnf_interface \
 218                     and metric_dict['metric_key'] == metric:
 219
 220                 self.monitor_lock.acquire()
 221
 222                 self.network_metrics.remove(metric_dict)
 223
 224                 #this removes the complete metric, all labels...
 225                 #REGISTRY.unregister(self.prom_metrics[metric_dict['metric_key']])
 226                 #self.registry.unregister(self.prom_metrics[metric_dict['metric_key']])
 227
 228                 for collector in self.registry._collectors :
 229                     #logging.info('name:{0} labels:{1} metrics:{2}'.format(collector._name, collector._labelnames, collector._metrics))
 230                     """
 231                     INFO:root:name:sonemu_rx_count_packets
 232                     labels:('vnf_name', 'vnf_interface')
 233                     metrics:{(u'tsrc', u'output'): < prometheus_client.core.Gauge
 234                     object
 235                     at
 236                     0x7f353447fd10 >}
 237                     """
 238                     logging.info('{0}'.format(collector._metrics.values()))
 239                     #if self.prom_metrics[metric_dict['metric_key']]
 240                     if (vnf_name, vnf_interface, 'None') in collector._metrics:
 241                         logging.info('2 name:{0} labels:{1} metrics:{2}'.format(collector._name, collector._labelnames,
 242                                                                               collector._metrics))
 243                         #collector._metrics = {}
 244                         collector.remove(vnf_name, vnf_interface, 'None')
 245
 246                 # set values to NaN, prometheus api currently does not support removal of metrics
 247                 #self.prom_metrics[metric_dict['metric_key']].labels(vnf_name, vnf_interface).set(float('nan'))
 248
 249                 # this removes the complete metric, all labels...
 250                 # 1 single monitor job for all metrics of the SDN controller
 251                 # we can only  remove from the pushgateway grouping keys(labels) which we have defined for the add_to_pushgateway
 252                 # we can not specify labels from the metrics to be removed
 253                 # if we need to remove the metrics seperatelty, we need to give them a separate grouping key, and probably a diffferent registry also
 254                 delete_from_gateway(self.pushgateway, job='sonemu-SDNcontroller')
 255
 256                 self.monitor_lock.release()
 257
 258                 logging.info('Stopped monitoring: {2} on {0}:{1}'.format(vnf_name, vnf_interface, metric))
 259                 return 'Stopped monitoring: {2} on {0}:{1}'.format(vnf_name, vnf_interface, metric)
 260
 261             # delete everything from this vnf
 262             elif metric_dict['vnf_name'] == vnf_name and vnf_interface is None and metric is None:
 263                 self.monitor_lock.acquire()
 264                 self.network_metrics.remove(metric_dict)
 265                 for collector in self.registry._collectors:
 266                     collector_dict = collector._metrics.copy()
 267                     for name, interface, id in collector_dict:
 268                         if name == vnf_name:
 269                             logging.info('3 name:{0} labels:{1} metrics:{2}'.format(collector._name, collector._labelnames,
 270                                                                            collector._metrics))
 271                             collector.remove(name, interface, 'None')
 272
 273                 delete_from_gateway(self.pushgateway, job='sonemu-SDNcontroller')
 274                 self.monitor_lock.release()
 275                 logging.info('Stopped monitoring vnf: {0}'.format(vnf_name))
 276                 return 'Stopped monitoring: {0}'.format(vnf_name)
 277
 278
 279     # get all metrics defined in the list and export it to Prometheus
 280     def get_flow_metrics(self):
 281         while self.start_monitoring:
 282
 283             self.monitor_flow_lock.acquire()
 284
 285             for flow_dict in self.flow_metrics:
 286                 data = {}
 287
 288                 data['cookie'] = flow_dict['cookie']
 289
 290                 if 'tx' in flow_dict['metric_key']:
 291                     data['match'] = {'in_port':flow_dict['mon_port']}
 292                 elif 'rx' in flow_dict['metric_key']:
 293                     data['out_port'] = flow_dict['mon_port']
 294
 295
 296                 # query Ryu
 297                 ret = self.net.ryu_REST('stats/flow', dpid=flow_dict['switch_dpid'], data=data)
 298                 flow_stat_dict = ast.literal_eval(ret)
 299
 300                 logging.info('received flow stat:{0} '.format(flow_stat_dict))
 301                 self.set_flow_metric(flow_dict, flow_stat_dict)
 302
 303             self.monitor_flow_lock.release()
 304             time.sleep(1)
 305
 306     def get_network_metrics(self):
 307         while self.start_monitoring:
 308
 309             self.monitor_lock.acquire()
 310
 311             # group metrics by dpid to optimize the rest api calls
 312             dpid_list = [metric_dict['switch_dpid'] for metric_dict in self.network_metrics]
 313             dpid_set = set(dpid_list)
 314
 315             for dpid in dpid_set:
 316
 317                 # query Ryu
 318                 ret = self.net.ryu_REST('stats/port', dpid=dpid)
 319                 port_stat_dict = ast.literal_eval(ret)
 320
 321                 metric_list = [metric_dict for metric_dict in self.network_metrics
 322                                if int(metric_dict['switch_dpid'])==int(dpid)]
 323                 #logging.info('1set prom packets:{0} '.format(self.network_metrics))
 324                 for metric_dict in metric_list:
 325                     self.set_network_metric(metric_dict, port_stat_dict)
 326
 327             self.monitor_lock.release()
 328             time.sleep(1)
 329
 330     # add metric to the list to export to Prometheus, parse the Ryu port-stats reply
 331     def set_network_metric(self, metric_dict, port_stat_dict):
 332         # vnf tx is the datacenter switch rx and vice-versa
 333         metric_key = self.switch_tx_rx(metric_dict['metric_key'])
 334         switch_dpid = metric_dict['switch_dpid']
 335         vnf_name = metric_dict['vnf_name']
 336         vnf_interface = metric_dict['vnf_interface']
 337         previous_measurement = metric_dict['previous_measurement']
 338         previous_monitor_time = metric_dict['previous_monitor_time']
 339         mon_port = metric_dict['mon_port']
 340
 341         for port_stat in port_stat_dict[str(switch_dpid)]:
 342             if int(port_stat['port_no']) == int(mon_port):
 343                 port_uptime = port_stat['duration_sec'] + port_stat['duration_nsec'] * 10 ** (-9)
 344                 this_measurement = int(port_stat[metric_key])
 345                 #logging.info('set prom packets:{0} {1}:{2}'.format(this_measurement, vnf_name, vnf_interface))
 346
 347                 # set prometheus metric
 348                 self.prom_metrics[metric_dict['metric_key']].\
 349                     labels({'vnf_name': vnf_name, 'vnf_interface': vnf_interface, 'flow_id': None}).\
 350                     set(this_measurement)
 351                 #push_to_gateway(self.pushgateway, job='SDNcontroller',
 352                 #                grouping_key={'metric':metric_dict['metric_key']}, registry=self.registry)
 353
 354                 # 1 single monitor job for all metrics of the SDN controller
 355                 pushadd_to_gateway(self.pushgateway, job='sonemu-SDNcontroller', registry=self.registry)
 356
 357                 if previous_monitor_time <= 0 or previous_monitor_time >= port_uptime:
 358                     metric_dict['previous_measurement'] = int(port_stat[metric_key])
 359                     metric_dict['previous_monitor_time'] = port_uptime
 360                     # do first measurement
 361                     #logging.info('first measurement')
 362                     time.sleep(1)
 363                     self.monitor_lock.release()
 364
 365                     metric_rate = self.get_network_metrics()
 366                     return metric_rate
 367
 368                 else:
 369                     time_delta = (port_uptime - metric_dict['previous_monitor_time'])
 370                     metric_rate = (this_measurement - metric_dict['previous_measurement']) / float(time_delta)
 371                     #logging.info('metric: {0} rate:{1}'.format(metric_dict['metric_key'], metric_rate))
 372
 373                 metric_dict['previous_measurement'] = this_measurement
 374                 metric_dict['previous_monitor_time'] = port_uptime
 375                 return metric_rate
 376
 377         logging.exception('metric {0} not found on {1}:{2}'.format(metric_key, vnf_name, vnf_interface))
 378         return 'metric {0} not found on {1}:{2}'.format(metric_key, vnf_name, vnf_interface)
 379
 380     def set_flow_metric(self, metric_dict, flow_stat_dict):
 381         # vnf tx is the datacenter switch rx and vice-versa
 382         #metric_key = self.switch_tx_rx(metric_dict['metric_key'])
 383         metric_key = metric_dict['metric_key']
 384         switch_dpid = metric_dict['switch_dpid']
 385         vnf_name = metric_dict['vnf_name']
 386         vnf_interface = metric_dict['vnf_interface']
 387         previous_measurement = metric_dict['previous_measurement']
 388         previous_monitor_time = metric_dict['previous_monitor_time']
 389         cookie = metric_dict['cookie']
 390
 391         # TODO aggregate all found flow stats
 392         flow_stat = flow_stat_dict[str(switch_dpid)][0]
 393         if 'bytes' in  metric_key:
 394             counter = flow_stat['byte_count']
 395         elif 'packet' in metric_key:
 396             counter = flow_stat['packet_count']
 397
 398         flow_uptime = flow_stat['duration_sec'] + flow_stat['duration_nsec'] * 10 ** (-9)
 399
 400         self.prom_metrics[metric_dict['metric_key']]. \
 401             labels({'vnf_name': vnf_name, 'vnf_interface': vnf_interface, 'flow_id': cookie}). \
 402             set(counter)
 403         pushadd_to_gateway(self.pushgateway, job='sonemu-SDNcontroller', registry=self.registry)
 404
 405         #logging.exception('metric {0} not found on {1}:{2}'.format(metric_key, vnf_name, vnf_interface))
 406         #return 'metric {0} not found on {1}:{2}'.format(metric_key, vnf_name, vnf_interface)
 407
 408     def query_Prometheus(self, query):
 409         '''
 410         escaped_chars='{}[]'
 411         for old in escaped_chars:
 412             new = '\{0}'.format(old)
 413             query = query.replace(old, new)
 414         '''
 415         url = self.prometheus_REST_api + '/' + 'api/v1/query?query=' + query
 416         #logging.info('query:{0}'.format(url))
 417         req = urllib2.Request(url)
 418         ret = urllib2.urlopen(req).read()
 419         ret = ast.literal_eval(ret)
 420         if ret['status'] == 'success':
 421             #logging.info('return:{0}'.format(ret))
 422             try:
 423                 ret = ret['data']['result'][0]['value']
 424             except:
 425                 ret = None
 426         else:
 427             ret = None
 428         return ret
 429
 430     def start_Prometheus(self, port=9090):
 431         # prometheus.yml configuration file is located in the same directory as this file
 432         cmd = ["docker",
 433                "run",
 434                "--rm",
 435                "-p", "{0}:9090".format(port),
 436                "-v", "{0}/prometheus.yml:/etc/prometheus/prometheus.yml".format(os.path.dirname(os.path.abspath(__file__))),
 437                "-v", "{0}/profile.rules:/etc/prometheus/profile.rules".format(os.path.dirname(os.path.abspath(__file__))),
 438                "--name", "prometheus",
 439                "prom/prometheus"
 440                ]
 441         logging.info('Start Prometheus container {0}'.format(cmd))
 442         return Popen(cmd)
 443
 444     def start_PushGateway(self, port=9091):
 445         cmd = ["docker",
 446                "run",
 447                "-d",
 448                "-p", "{0}:9091".format(port),
 449                "--name", "pushgateway",
 450                "prom/pushgateway"
 451                ]
 452
 453         logging.info('Start Prometheus Push Gateway container {0}'.format(cmd))
 454         return Popen(cmd)
 455
 456     def start_cadvisor(self, port=8090):
 457         cmd = ["docker",
 458                "run",
 459                "--rm",
 460                "--volume=/:/rootfs:ro",
 461                "--volume=/var/run:/var/run:rw",
 462                "--volume=/sys:/sys:ro",
 463                "--volume=/var/lib/docker/:/var/lib/docker:ro",
 464                "--publish={0}:8080".format(port),
 465                "--name=cadvisor",
 466                "google/cadvisor:latest"
 467                ]
 468         logging.info('Start cAdvisor container {0}'.format(cmd))
 469         return Popen(cmd)
 470
 471     def stop(self):
 472         # stop the monitoring thread
 473         self.start_monitoring = False
 474         self.monitor_thread.join()
 475         self.monitor_flow_thread.join()
 476
 477         if self.prometheus_process is not None:
 478             logging.info('stopping prometheus container')
 479             self.prometheus_process.terminate()
 480             self.prometheus_process.kill()
 481             self._stop_container('prometheus')
 482
 483         if self.pushgateway_process is not None:
 484             logging.info('stopping pushgateway container')
 485             self.pushgateway_process.terminate()
 486             self.pushgateway_process.kill()
 487             self._stop_container('pushgateway')
 488
 489         if self.cadvisor_process is not None:
 490             logging.info('stopping cadvisor container')
 491             self.cadvisor_process.terminate()
 492             self.cadvisor_process.kill()
 493             self._stop_container('cadvisor')
 494
 495     def switch_tx_rx(self,metric=''):
 496         # when monitoring vnfs, the tx of the datacenter switch is actually the rx of the vnf
 497         # so we need to change the metric name to be consistent with the vnf rx or tx
 498         if 'tx' in metric:
 499             metric = metric.replace('tx','rx')
 500         elif 'rx' in metric:
 501             metric = metric.replace('rx','tx')
 502
 503         return metric
 504
 505     def _stop_container(self, name):
 506         cmd = ["docker",
 507                "stop",
 508                name]
 509         Popen(cmd).wait()
 510
 511         cmd = ["docker",
 512                "rm",
 513                name]
 514         Popen(cmd).wait()
 515
 516     def profile(self, mgmt_ip, rate, input_ip, vnf_uuid ):
 517
 518         ssh = paramiko.SSHClient()
 519         ssh.set_missing_host_key_policy(paramiko.AutoAddPolicy())
 520         #ssh.connect(mgmt_ip, username='steven', password='test')
 521         ssh.connect(mgmt_ip, username='root', password='root')
 522
 523         iperf_cmd = 'iperf -c {0} -u -l18 -b{1}M -t1000 &'.format(input_ip, rate)
 524         if rate > 0:
 525             stdin, stdout, stderr = ssh.exec_command(iperf_cmd)
 526
 527         start_time = time.time()
 528         query_cpu = '(sum(rate(container_cpu_usage_seconds_total{{id="/docker/{0}"}}[{1}s])))'.format(vnf_uuid, 1)
 529         while (time.time() - start_time) < 15:
 530             data = self.query_Prometheus(query_cpu)
 531             # logging.info('rate: {1} data:{0}'.format(data, rate))
 532             gevent.sleep(0)
 533             time.sleep(1)
 534
 535         query_cpu2 = '(sum(rate(container_cpu_usage_seconds_total{{id="/docker/{0}"}}[{1}s])))'.format(vnf_uuid, 8)
 536         cpu_load = float(self.query_Prometheus(query_cpu2)[1])
 537         output = 'rate: {1}Mbps; cpu_load: {0}%'.format(round(cpu_load * 100, 2), rate)
 538         output_line = output
 539         logging.info(output_line)
 540
 541         stop_iperf = 'pkill -9 iperf'
 542         stdin, stdout, stderr = ssh.exec_command(stop_iperf)
 543
 544         return output_line
 545