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 logging.basicConfig(level=logging.INFO)
  15
  16 """
  17 class to read openflow stats from the Ryu controller of the DCNetwork
  18 """
  19
  20 class DCNetworkMonitor():
  21     def __init__(self, net):
  22         self.net = net
  23         # link to Ryu REST_API
  24         self.ip = '0.0.0.0'
  25         self.port = '8080'
  26         self.REST_api = 'http://{0}:{1}'.format(self.ip,self.port)
  27
  28         # helper variables to calculate the metrics
  29         self.pushgateway = 'localhost:9091'
  30         # Start up the server to expose the metrics to Prometheus.
  31         #start_http_server(8000)
  32         # supported Prometheus metrics
  33         self.registry = CollectorRegistry()
  34         self.prom_tx_packet_count = Gauge('sonemu_tx_count_packets', 'Total number of packets sent',
  35                                           ['vnf_name', 'vnf_interface'], registry=self.registry)
  36         self.prom_rx_packet_count = Gauge('sonemu_rx_count_packets', 'Total number of packets received',
  37                                           ['vnf_name', 'vnf_interface'], registry=self.registry)
  38         self.prom_tx_byte_count = Gauge('sonemu_tx_count_bytes', 'Total number of bytes sent',
  39                                         ['vnf_name', 'vnf_interface'], registry=self.registry)
  40         self.prom_rx_byte_count = Gauge('sonemu_rx_count_bytes', 'Total number of bytes received',
  41                                         ['vnf_name', 'vnf_interface'], registry=self.registry)
  42
  43         self.prom_metrics={'tx_packets':self.prom_tx_packet_count, 'rx_packets':self.prom_rx_packet_count,
  44                            'tx_bytes':self.prom_tx_byte_count,'rx_bytes':self.prom_rx_byte_count}
  45
  46         # list of installed metrics to monitor
  47         # each entry can contain this data
  48         '''
  49         {
  50         switch_dpid = 0
  51         vnf_name = None
  52         vnf_interface = None
  53         previous_measurement = 0
  54         previous_monitor_time = 0
  55         metric_key = None
  56         mon_port = None
  57         }
  58         '''
  59         self.monitor_lock = threading.Lock()
  60         self.network_metrics = []
  61
  62         # start monitoring thread
  63         self.start_monitoring = True
  64         self.monitor_thread = threading.Thread(target=self.get_network_metrics)
  65         self.monitor_thread.start()
  66
  67         # helper tools
  68         self.pushgateway_process = self.start_PushGateway()
  69         self.prometheus_process = self.start_Prometheus()
  70         self.cadvisor_process = self.start_cadvisor()
  71
  72     # first set some parameters, before measurement can start
  73     def setup_metric(self, vnf_name, vnf_interface=None, metric='tx_packets'):
  74
  75         network_metric = {}
  76
  77         # check if port is specified (vnf:port)
  78         if vnf_interface is None:
  79             # take first interface by default
  80             connected_sw = self.net.DCNetwork_graph.neighbors(vnf_name)[0]
  81             link_dict = self.net.DCNetwork_graph[vnf_name][connected_sw]
  82             vnf_interface = link_dict[0]['src_port_id']
  83
  84         network_metric['vnf_name'] = vnf_name
  85         network_metric['vnf_interface'] = vnf_interface
  86
  87         for connected_sw in self.net.DCNetwork_graph.neighbors(vnf_name):
  88             link_dict = self.net.DCNetwork_graph[vnf_name][connected_sw]
  89             for link in link_dict:
  90                 # logging.info("{0},{1}".format(link_dict[link],vnf_interface))
  91                 if link_dict[link]['src_port_id'] == vnf_interface:
  92                     # found the right link and connected switch
  93                     # logging.info("{0},{1}".format(link_dict[link]['src_port_id'], vnf_source_interface))
  94                     network_metric['mon_port'] = link_dict[link]['dst_port']
  95                     break
  96
  97         if 'mon_port' not in network_metric:
  98             logging.exception("vnf interface {0}:{1} not found!".format(vnf_name,vnf_interface))
  99             return "vnf interface {0}:{1} not found!".format(vnf_name,vnf_interface)
 100
 101         try:
 102             # default port direction to monitor
 103             if metric is None:
 104                 metric = 'tx_packets'
 105
 106             vnf_switch = self.net.DCNetwork_graph.neighbors(str(vnf_name))
 107
 108             if len(vnf_switch) > 1:
 109                 logging.info("vnf: {0} has multiple ports".format(vnf_name))
 110                 return
 111             elif len(vnf_switch) == 0:
 112                 logging.info("vnf: {0} is not connected".format(vnf_name))
 113                 return
 114             else:
 115                 vnf_switch = vnf_switch[0]
 116             next_node = self.net.getNodeByName(vnf_switch)
 117
 118             if not isinstance(next_node, OVSSwitch):
 119                 logging.info("vnf: {0} is not connected to switch".format(vnf_name))
 120                 return
 121
 122             network_metric['previous_measurement'] = 0
 123             network_metric['previous_monitor_time'] = 0
 124
 125
 126             network_metric['switch_dpid'] = int(str(next_node.dpid), 16)
 127             network_metric['metric_key'] = metric
 128
 129             self.monitor_lock.acquire()
 130
 131             self.network_metrics.append(network_metric)
 132             self.monitor_lock.release()
 133
 134
 135             logging.info('Started monitoring: {2} on {0}:{1}'.format(vnf_name, vnf_interface, metric))
 136             return 'Started monitoring: {2} on {0}:{1}'.format(vnf_name, vnf_interface, metric)
 137
 138         except Exception as ex:
 139             logging.exception("setup_metric error.")
 140             return ex.message
 141
 142     def stop_metric(self, vnf_name, vnf_interface, metric):
 143
 144         for metric_dict in self.network_metrics:
 145             if metric_dict['vnf_name'] == vnf_name and metric_dict['vnf_interface'] == vnf_interface \
 146                     and metric_dict['metric_key'] == metric:
 147
 148                 self.monitor_lock.acquire()
 149
 150                 self.network_metrics.remove(metric_dict)
 151
 152                 #this removes the complete metric, all labels...
 153                 #REGISTRY.unregister(self.prom_metrics[metric_dict['metric_key']])
 154                 #self.registry.unregister(self.prom_metrics[metric_dict['metric_key']])
 155
 156                 for collector in self.registry._collectors :
 157                     logging.info('name:{0} labels:{1} metrics:{2}'.format(collector._name, collector._labelnames, collector._metrics))
 158                     """
 159                     INFO:root:name:sonemu_rx_count_packets
 160                     labels:('vnf_name', 'vnf_interface')
 161                     metrics:{(u'tsrc', u'output'): < prometheus_client.core.Gauge
 162                     object
 163                     at
 164                     0x7f353447fd10 >}
 165                     """
 166                     logging.info('{0}'.format(collector._metrics.values()))
 167                     #if self.prom_metrics[metric_dict['metric_key']]
 168                     if (vnf_name, vnf_interface) in collector._metrics:
 169                         logging.info('2 name:{0} labels:{1} metrics:{2}'.format(collector._name, collector._labelnames,
 170                                                                               collector._metrics))
 171                         #collector._metrics = {}
 172                         collector.remove(vnf_name, vnf_interface)
 173
 174                 # set values to NaN, prometheus api currently does not support removal of metrics
 175                 #self.prom_metrics[metric_dict['metric_key']].labels(vnf_name, vnf_interface).set(float('nan'))
 176
 177                 # this removes the complete metric, all labels...
 178                 # 1 single monitor job for all metrics of the SDN controller
 179                 # we can only  remove from the pushgateway grouping keys(labels) which we have defined for the add_to_pushgateway
 180                 # we can not specify labels from the metrics to be removed
 181                 # if we need to remove the metrics seperatelty, we need to give them a separate grouping key, and probably a diffferent registry also
 182                 delete_from_gateway(self.pushgateway, job='sonemu-SDNcontroller')
 183
 184                 self.monitor_lock.release()
 185
 186                 logging.info('Stopped monitoring: {2} on {0}:{1}'.format(vnf_name, vnf_interface, metric))
 187                 return 'Stopped monitoring: {2} on {0}:{1}'.format(vnf_name, vnf_interface, metric)
 188
 189
 190     # get all metrics defined in the list and export it to Prometheus
 191     def get_network_metrics(self):
 192         while self.start_monitoring:
 193
 194             self.monitor_lock.acquire()
 195
 196             # group metrics by dpid to optimize the rest api calls
 197             dpid_list = [metric_dict['switch_dpid'] for metric_dict in self.network_metrics]
 198             dpid_set = set(dpid_list)
 199
 200             for dpid in dpid_set:
 201
 202                 # query Ryu
 203                 ret = self.REST_cmd('stats/port', dpid)
 204                 port_stat_dict = ast.literal_eval(ret)
 205
 206                 metric_list = [metric_dict for metric_dict in self.network_metrics
 207                                if int(metric_dict['switch_dpid'])==int(dpid)]
 208                 #logging.info('1set prom packets:{0} '.format(self.network_metrics))
 209                 for metric_dict in metric_list:
 210                     self.set_network_metric(metric_dict, port_stat_dict)
 211
 212             self.monitor_lock.release()
 213             time.sleep(1)
 214
 215     # add metric to the list to export to Prometheus, parse the Ryu port-stats reply
 216     def set_network_metric(self, metric_dict, port_stat_dict):
 217         # vnf tx is the datacenter switch rx and vice-versa
 218         metric_key = self.switch_tx_rx(metric_dict['metric_key'])
 219         switch_dpid = metric_dict['switch_dpid']
 220         vnf_name = metric_dict['vnf_name']
 221         vnf_interface = metric_dict['vnf_interface']
 222         previous_measurement = metric_dict['previous_measurement']
 223         previous_monitor_time = metric_dict['previous_monitor_time']
 224         mon_port = metric_dict['mon_port']
 225
 226         for port_stat in port_stat_dict[str(switch_dpid)]:
 227             if int(port_stat['port_no']) == int(mon_port):
 228                 port_uptime = port_stat['duration_sec'] + port_stat['duration_nsec'] * 10 ** (-9)
 229                 this_measurement = int(port_stat[metric_key])
 230                 #logging.info('set prom packets:{0} {1}:{2}'.format(this_measurement, vnf_name, vnf_interface))
 231
 232                 # set prometheus metric
 233                 self.prom_metrics[metric_dict['metric_key']].\
 234                     labels({'vnf_name':vnf_name, 'vnf_interface':vnf_interface}).\
 235                     set(this_measurement)
 236                 #push_to_gateway(self.pushgateway, job='SDNcontroller',
 237                 #                grouping_key={'metric':metric_dict['metric_key']}, registry=self.registry)
 238
 239                 # 1 single monitor job for all metrics of the SDN controller
 240                 pushadd_to_gateway(self.pushgateway, job='sonemu-SDNcontroller', registry=self.registry)
 241
 242                 if previous_monitor_time <= 0 or previous_monitor_time >= port_uptime:
 243                     metric_dict['previous_measurement'] = int(port_stat[metric_key])
 244                     metric_dict['previous_monitor_time'] = port_uptime
 245                     # do first measurement
 246                     #logging.info('first measurement')
 247                     time.sleep(1)
 248                     self.monitor_lock.release()
 249                     metric_rate = self.get_network_metrics()
 250                     return metric_rate
 251                 else:
 252                     time_delta = (port_uptime - metric_dict['previous_monitor_time'])
 253                     metric_rate = (this_measurement - metric_dict['previous_measurement']) / float(time_delta)
 254                     # logging.info('uptime:{2} delta:{0} rate:{1}'.format(time_delta,byte_rate,port_uptime))
 255
 256                 metric_dict['previous_measurement'] = this_measurement
 257                 metric_dict['previous_monitor_time'] = port_uptime
 258                 return metric_rate
 259
 260         logging.exception('metric {0} not found on {1}:{2}'.format(metric_key, vnf_name, vnf_interface))
 261         return 'metric {0} not found on {1}:{2}'.format(metric_key, vnf_name, vnf_interface)
 262
 263
 264     def REST_cmd(self, prefix, dpid):
 265         url = self.REST_api + '/' + str(prefix) + '/' + str(dpid)
 266         req = urllib2.Request(url)
 267         ret = urllib2.urlopen(req).read()
 268         return ret
 269
 270     def start_Prometheus(self, port=9090):
 271         # prometheus.yml configuration file is located in the same directory as this file
 272         cmd = ["docker",
 273                "run",
 274                "--rm",
 275                "-p", "{0}:9090".format(port),
 276                "-v", "{0}/prometheus.yml:/etc/prometheus/prometheus.yml".format(os.path.dirname(os.path.abspath(__file__))),
 277                "--name", "prometheus",
 278                "prom/prometheus"
 279                ]
 280         logging.info('Start Prometheus container {0}'.format(cmd))
 281         return Popen(cmd)
 282
 283     def start_PushGateway(self, port=9091):
 284         cmd = ["docker",
 285                "run",
 286                "-d",
 287                "-p", "{0}:9091".format(port),
 288                "--name", "pushgateway",
 289                "prom/pushgateway"
 290                ]
 291
 292         logging.info('Start Prometheus Push Gateway container {0}'.format(cmd))
 293         return Popen(cmd)
 294
 295     def start_cadvisor(self, port=8090):
 296         cmd = ["docker",
 297                "run",
 298                "--rm",
 299                "--volume=/:/rootfs:ro",
 300                "--volume=/var/run:/var/run:rw",
 301                "--volume=/sys:/sys:ro",
 302                "--volume=/var/lib/docker/:/var/lib/docker:ro",
 303                "--publish={0}:8080".format(port),
 304                "--name=cadvisor",
 305                "google/cadvisor:latest"
 306                ]
 307         logging.info('Start cAdvisor container {0}'.format(cmd))
 308         return Popen(cmd)
 309
 310     def stop(self):
 311         # stop the monitoring thread
 312         self.start_monitoring = False
 313         self.monitor_thread.join()
 314
 315         if self.prometheus_process is not None:
 316             logging.info('stopping prometheus container')
 317             self.prometheus_process.terminate()
 318             self.prometheus_process.kill()
 319             self._stop_container('prometheus')
 320
 321         if self.pushgateway_process is not None:
 322             logging.info('stopping pushgateway container')
 323             self.pushgateway_process.terminate()
 324             self.pushgateway_process.kill()
 325             self._stop_container('pushgateway')
 326
 327         if self.cadvisor_process is not None:
 328             logging.info('stopping cadvisor container')
 329             self.cadvisor_process.terminate()
 330             self.cadvisor_process.kill()
 331             self._stop_container('cadvisor')
 332
 333     def switch_tx_rx(self,metric=''):
 334         # when monitoring vnfs, the tx of the datacenter switch is actually the rx of the vnf
 335         # so we need to change the metric name to be consistent with the vnf rx or tx
 336         if 'tx' in metric:
 337             metric = metric.replace('tx','rx')
 338         elif 'rx' in metric:
 339             metric = metric.replace('rx','tx')
 340
 341         return metric
 342
 343     def _stop_container(self, name):
 344         cmd = ["docker",
 345                "stop",
 346                name]
 347         Popen(cmd).wait()
 348
 349         cmd = ["docker",
 350                "rm",
 351                name]
 352         Popen(cmd).wait()
 353
 354