+ if not isinstance(next_node, OVSSwitch):\r
+ logging.info(\r
+ "vnf: {0} is not connected to switch".format(vnf_name))\r
+ return\r
+\r
+ network_metric['previous_measurement'] = 0\r
+ network_metric['previous_monitor_time'] = 0\r
+\r
+ network_metric['switch_dpid'] = int(str(next_node.dpid), 16)\r
+ network_metric['metric_key'] = metric\r
+\r
+ self.monitor_lock.acquire()\r
+ self.network_metrics.append(network_metric)\r
+ self.monitor_lock.release()\r
+\r
+ logging.info('Started monitoring: {2} on {0}:{1}'.format(\r
+ vnf_name, vnf_interface, metric))\r
+ return 'Started monitoring: {2} on {0}:{1}'.format(\r
+ vnf_name, vnf_interface, metric)\r
+\r
+ except Exception as ex:\r
+ logging.exception("setup_metric error.")\r
+ return ex.message\r
+\r
+ def stop_metric(self, vnf_name, vnf_interface=None, metric=None):\r
+\r
+ # check if port is specified (vnf:port)\r
+ if vnf_interface is None and metric is not None:\r
+ # take first interface by default\r
+ connected_sw = self.net.DCNetwork_graph.neighbors(vnf_name)[0]\r
+ link_dict = self.net.DCNetwork_graph[vnf_name][connected_sw]\r
+ vnf_interface = link_dict[0]['src_port_id']\r
+\r
+ for metric_dict in deepcopy(self.network_metrics):\r
+ if metric_dict['vnf_name'] == vnf_name and metric_dict['vnf_interface'] == vnf_interface \\r
+ and metric_dict['metric_key'] == metric:\r
+\r
+ self.monitor_lock.acquire()\r
+\r
+ self.network_metrics.remove(metric_dict)\r
+\r
+ # set values to NaN, prometheus api currently does not support removal of metrics\r
+ # self.prom_metrics[metric_dict['metric_key']].labels(vnf_name, vnf_interface).set(float('nan'))\r
+ self.prom_metrics[metric_dict['metric_key']]. \\r
+ labels(vnf_name=vnf_name, vnf_interface=vnf_interface, flow_id=None). \\r
+ set(float('nan'))\r
+\r
+ # this removes the complete metric, all labels...\r
+ # 1 single monitor job for all metrics of the SDN controller\r
+ # we can only remove from the pushgateway grouping keys(labels) which we have defined for the add_to_pushgateway\r
+ # we can not specify labels from the metrics to be removed\r
+ # if we need to remove the metrics seperatelty, we need to give\r
+ # them a separate grouping key, and probably a diffferent\r
+ # registry also\r
+ delete_from_gateway(\r
+ self.pushgateway, job='sonemu-SDNcontroller')\r
+\r
+ self.monitor_lock.release()\r
+\r
+ logging.info('Stopped monitoring: {2} on {0}:{1}'.format(\r
+ vnf_name, vnf_interface, metric))\r
+ return 'Stopped monitoring: {2} on {0}:{1}'.format(\r
+ vnf_name, vnf_interface, metric)\r
+\r
+ # delete everything from this vnf\r
+ elif metric_dict['vnf_name'] == vnf_name and vnf_interface is None and metric is None:\r
+ self.monitor_lock.acquire()\r
+ self.network_metrics.remove(metric_dict)\r
+ logging.info('remove metric from monitor: vnf_name:{0} vnf_interface:{1} mon_port:{2}'.format(\r
+ metric_dict['vnf_name'], metric_dict['vnf_interface'], metric_dict['mon_port']))\r
+\r
+ delete_from_gateway(\r
+ self.pushgateway, job='sonemu-SDNcontroller')\r
+ self.monitor_lock.release()\r
+ continue\r
+\r
+ if vnf_interface is None and metric is None:\r
+ logging.info('Stopped monitoring vnf: {0}'.format(vnf_name))\r
+ return 'Stopped monitoring: {0}'.format(vnf_name)\r
+ else:\r
+ return 'Error stopping monitoring metric: {0} on {1}:{2}'.format(\r
+ metric, vnf_name, vnf_interface)\r
+\r
+ def get_flow_metrics(self):\r
+ """\r
+ Get all metrics defined in the list and export it to Prometheus.\r
+ """\r
+ while self.start_monitoring:\r
+\r
+ self.monitor_flow_lock.acquire()\r
+\r
+ for flow_dict in self.flow_metrics:\r
+ data = {}\r
+\r
+ data['cookie'] = flow_dict['cookie']\r
+ data['cookie_mask'] = COOKIE_MASK\r
+\r
+ if 'tx' in flow_dict['metric_key']:\r
+ data['match'] = {'in_port': flow_dict['mon_port']}\r
+ elif 'rx' in flow_dict['metric_key']:\r
+ data['out_port'] = flow_dict['mon_port']\r
+\r
+ # query Ryu\r
+ ret = self.net.ryu_REST(\r
+ 'stats/flow', dpid=flow_dict['switch_dpid'], data=data)\r
+ if isinstance(ret, dict):\r
+ flow_stat_dict = ret\r
+ elif isinstance(ret, basestring):\r
+ flow_stat_dict = ast.literal_eval(ret.rstrip())\r
+ else:\r
+ flow_stat_dict = None\r
+\r
+ logging.debug('received flow stat:{0} '.format(flow_stat_dict))\r
+\r
+ self.set_flow_metric(flow_dict, flow_stat_dict)\r
+\r
+ try:\r
+ if len(self.flow_metrics) > 0:\r
+ pushadd_to_gateway(\r
+ self.pushgateway, job='sonemu-SDNcontroller', registry=self.registry)\r
+ except Exception as e:\r
+ logging.warning(\r
+ "Pushgateway not reachable: {0} {1}".format(Exception, e))\r
+\r
+ self.monitor_flow_lock.release()\r
+ time.sleep(1)\r
+\r
+ def get_network_metrics(self):\r
+ while self.start_monitoring:\r
+\r
+ self.monitor_lock.acquire()\r
+\r
+ # group metrics by dpid to optimize the rest api calls\r
+ dpid_list = [metric_dict['switch_dpid']\r
+ for metric_dict in self.network_metrics]\r
+ dpid_set = set(dpid_list)\r
+\r
+ for dpid in dpid_set:\r
+\r
+ # query Ryu\r
+ ret = self.net.ryu_REST('stats/port', dpid=dpid)\r
+ if isinstance(ret, dict):\r
+ port_stat_dict = ret\r
+ elif isinstance(ret, basestring):\r
+ port_stat_dict = ast.literal_eval(ret.rstrip())\r
+ else:\r
+ port_stat_dict = None\r
+\r
+ metric_list = [metric_dict for metric_dict in self.network_metrics\r
+ if int(metric_dict['switch_dpid']) == int(dpid)]\r
+\r
+ for metric_dict in metric_list:\r
+ self.set_network_metric(metric_dict, port_stat_dict)\r
+\r
+ try:\r
+ if len(self.network_metrics) > 0:\r
+ pushadd_to_gateway(\r
+ self.pushgateway, job='sonemu-SDNcontroller', registry=self.registry)\r
+ except Exception as e:\r
+ logging.warning(\r
+ "Pushgateway not reachable: {0} {1}".format(Exception, e))\r
+\r
+ self.monitor_lock.release()\r
+ time.sleep(1)\r
+\r
+ # add metric to the list to export to Prometheus, parse the Ryu port-stats\r
+ # reply\r
+ def set_network_metric(self, metric_dict, port_stat_dict):\r
+ # vnf tx is the datacenter switch rx and vice-versa\r
+ metric_key = self.switch_tx_rx(metric_dict['metric_key'])\r
+ switch_dpid = metric_dict['switch_dpid']\r
+ vnf_name = metric_dict['vnf_name']\r
+ vnf_interface = metric_dict['vnf_interface']\r
+ previous_monitor_time = metric_dict['previous_monitor_time']\r
+ mon_port = metric_dict['mon_port']\r
+ for port_stat in port_stat_dict[str(switch_dpid)]:\r
+ # ovs output also gives back 'LOCAL' port\r
+ if port_stat['port_no'] == 'LOCAL':\r
+ continue\r
+ if int(port_stat['port_no']) == int(mon_port):\r
+ port_uptime = port_stat['duration_sec'] + \\r
+ port_stat['duration_nsec'] * 10 ** (-9)\r
+ this_measurement = int(port_stat[metric_key])\r
+\r
+ # set prometheus metric\r
+ self.prom_metrics[metric_dict['metric_key']].\\r
+ labels(vnf_name=vnf_name, vnf_interface=vnf_interface, flow_id=None).\\r
+ set(this_measurement)\r
+\r
+ # also the rate is calculated here, but not used for now\r
+ # (rate can be easily queried from prometheus also)\r
+ if previous_monitor_time <= 0 or previous_monitor_time >= port_uptime:\r
+ metric_dict['previous_measurement'] = int(\r
+ port_stat[metric_key])\r
+ metric_dict['previous_monitor_time'] = port_uptime\r
+ # do first measurement\r
+ # time.sleep(1)\r
+ # self.monitor_lock.release()\r
+ # rate cannot be calculated yet (need a first measurement)\r
+ metric_dict['previous_measurement'] = this_measurement\r
+ metric_dict['previous_monitor_time'] = port_uptime\r