-__author__ = 'Administrator'\r
+"""\r
+Copyright (c) 2015 SONATA-NFV\r
+ALL RIGHTS RESERVED.\r
+\r
+Licensed under the Apache License, Version 2.0 (the "License");\r
+you may not use this file except in compliance with the License.\r
+You may obtain a copy of the License at\r
+\r
+ http://www.apache.org/licenses/LICENSE-2.0\r
+\r
+Unless required by applicable law or agreed to in writing, software\r
+distributed under the License is distributed on an "AS IS" BASIS,\r
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\r
+See the License for the specific language governing permissions and\r
+limitations under the License.\r
+\r
+Neither the name of the SONATA-NFV [, ANY ADDITIONAL AFFILIATION]\r
+nor the names of its contributors may be used to endorse or promote\r
+products derived from this software without specific prior written\r
+permission.\r
+\r
+This work has been performed in the framework of the SONATA project,\r
+funded by the European Commission under Grant number 671517 through\r
+the Horizon 2020 and 5G-PPP programmes. The authors would like to\r
+acknowledge the contributions of their colleagues of the SONATA\r
+partner consortium (www.sonata-nfv.eu).\r
+"""\r
\r
-import urllib2\r
import logging\r
+import sys\r
from mininet.node import OVSSwitch\r
import ast\r
import time\r
-from prometheus_client import start_http_server, Summary, Histogram, Gauge, Counter, REGISTRY\r
+from prometheus_client import start_http_server, Summary, Histogram, Gauge, Counter, REGISTRY, CollectorRegistry, \\r
+ pushadd_to_gateway, push_to_gateway, delete_from_gateway\r
import threading\r
-from subprocess import Popen, PIPE\r
+from subprocess import Popen\r
import os\r
\r
+\r
logging.basicConfig(level=logging.INFO)\r
\r
"""\r
class to read openflow stats from the Ryu controller of the DCNetwork\r
"""\r
\r
+PUSHGATEWAY_PORT = 9091\r
+# we cannot use port 8080 because ryu-ofrest api is already using that one\r
+CADVISOR_PORT = 8081\r
+\r
+COOKIE_MASK = 0xffffffff\r
+\r
class DCNetworkMonitor():\r
def __init__(self, net):\r
self.net = net\r
- # link to Ryu REST_API\r
- self.ip = '0.0.0.0'\r
- self.port = '8080'\r
- self.REST_api = 'http://{0}:{1}'.format(self.ip,self.port)\r
-\r
- # helper variables to calculate the metrics\r
- # Start up the server to expose the metrics to Prometheus.\r
- start_http_server(8000)\r
+\r
+ # pushgateway address\r
+ self.pushgateway = 'localhost:{0}'.format(PUSHGATEWAY_PORT)\r
+\r
# supported Prometheus metrics\r
+ self.registry = CollectorRegistry()\r
self.prom_tx_packet_count = Gauge('sonemu_tx_count_packets', 'Total number of packets sent',\r
- ['vnf_name', 'vnf_interface'])\r
+ ['vnf_name', 'vnf_interface', 'flow_id'], registry=self.registry)\r
self.prom_rx_packet_count = Gauge('sonemu_rx_count_packets', 'Total number of packets received',\r
- ['vnf_name', 'vnf_interface'])\r
+ ['vnf_name', 'vnf_interface', 'flow_id'], registry=self.registry)\r
self.prom_tx_byte_count = Gauge('sonemu_tx_count_bytes', 'Total number of bytes sent',\r
- ['vnf_name', 'vnf_interface'])\r
+ ['vnf_name', 'vnf_interface', 'flow_id'], registry=self.registry)\r
self.prom_rx_byte_count = Gauge('sonemu_rx_count_bytes', 'Total number of bytes received',\r
- ['vnf_name', 'vnf_interface'])\r
+ ['vnf_name', 'vnf_interface', 'flow_id'], registry=self.registry)\r
\r
self.prom_metrics={'tx_packets':self.prom_tx_packet_count, 'rx_packets':self.prom_rx_packet_count,\r
'tx_bytes':self.prom_tx_byte_count,'rx_bytes':self.prom_rx_byte_count}\r
mon_port = None\r
}\r
'''\r
+ self.monitor_lock = threading.Lock()\r
+ self.monitor_flow_lock = threading.Lock()\r
self.network_metrics = []\r
+ self.flow_metrics = []\r
\r
# start monitoring thread\r
self.start_monitoring = True\r
self.monitor_thread = threading.Thread(target=self.get_network_metrics)\r
self.monitor_thread.start()\r
\r
+ self.monitor_flow_thread = threading.Thread(target=self.get_flow_metrics)\r
+ self.monitor_flow_thread.start()\r
+\r
# helper tools\r
- self.prometheus_process = self.start_Prometheus()\r
- self.cadvisor_process = self.start_cadvisor()\r
+ # cAdvisor, Prometheus pushgateway are started as external container, to gather monitoring metric in son-emu\r
+ self.pushgateway_process = self.start_PushGateway()\r
+ self.cadvisor_process = self.start_cAdvisor()\r
+\r
+\r
+ # first set some parameters, before measurement can start\r
+ def setup_flow(self, vnf_name, vnf_interface=None, metric='tx_packets', cookie=0):\r
+\r
+ flow_metric = {}\r
+\r
+ # check if port is specified (vnf:port)\r
+ if vnf_interface is 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
+ flow_metric['vnf_name'] = vnf_name\r
+ flow_metric['vnf_interface'] = vnf_interface\r
+\r
+ vnf_switch = None\r
+ for connected_sw in self.net.DCNetwork_graph.neighbors(vnf_name):\r
+ link_dict = self.net.DCNetwork_graph[vnf_name][connected_sw]\r
+ for link in link_dict:\r
+ if link_dict[link]['src_port_id'] == vnf_interface:\r
+ # found the right link and connected switch\r
+ vnf_switch = connected_sw\r
+ flow_metric['mon_port'] = link_dict[link]['dst_port_nr']\r
+ break\r
+\r
+ if not vnf_switch:\r
+ logging.exception("vnf switch of {0}:{1} not found!".format(vnf_name, vnf_interface))\r
+ return "vnf switch of {0}:{1} not found!".format(vnf_name, vnf_interface)\r
+\r
+ try:\r
+ # default port direction to monitor\r
+ if metric is None:\r
+ metric = 'tx_packets'\r
+\r
+ next_node = self.net.getNodeByName(vnf_switch)\r
+\r
+ if not isinstance(next_node, OVSSwitch):\r
+ logging.info("vnf: {0} is not connected to switch".format(vnf_name))\r
+ return\r
+\r
+ flow_metric['previous_measurement'] = 0\r
+ flow_metric['previous_monitor_time'] = 0\r
+\r
+ flow_metric['switch_dpid'] = int(str(next_node.dpid), 16)\r
+ flow_metric['metric_key'] = metric\r
+ flow_metric['cookie'] = cookie\r
+\r
+ self.monitor_flow_lock.acquire()\r
+ self.flow_metrics.append(flow_metric)\r
+ self.monitor_flow_lock.release()\r
+\r
+ logging.info('Started monitoring flow:{3} {2} on {0}:{1}'.format(vnf_name, vnf_interface, metric, cookie))\r
+ return 'Started monitoring flow:{3} {2} on {0}:{1}'.format(vnf_name, vnf_interface, metric, cookie)\r
+\r
+ except Exception as ex:\r
+ logging.exception("setup_metric error.")\r
+ return ex.message\r
+\r
+ def stop_flow(self, vnf_name, vnf_interface=None, metric=None, cookie=0,):\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 flow_dict in self.flow_metrics:\r
+ if flow_dict['vnf_name'] == vnf_name and flow_dict['vnf_interface'] == vnf_interface \\r
+ and flow_dict['metric_key'] == metric and flow_dict['cookie'] == cookie:\r
+\r
+ self.monitor_flow_lock.acquire()\r
+\r
+ self.flow_metrics.remove(flow_dict)\r
+\r
+ for collector in self.registry._collectors:\r
+ if (vnf_name, vnf_interface, cookie) in collector._metrics:\r
+ collector.remove(vnf_name, vnf_interface, cookie)\r
+\r
+ delete_from_gateway(self.pushgateway, job='sonemu-SDNcontroller')\r
+\r
+ self.monitor_flow_lock.release()\r
+\r
+ logging.info('Stopped monitoring flow {3}: {2} on {0}:{1}'.format(vnf_name, vnf_interface, metric, cookie))\r
+ return 'Stopped monitoring flow {3}: {2} on {0}:{1}'.format(vnf_name, vnf_interface, metric, cookie)\r
+\r
+ return 'Error stopping monitoring flow: {0} on {1}:{2}'.format(metric, vnf_name, vnf_interface)\r
+\r
\r
# first set some parameters, before measurement can start\r
def setup_metric(self, vnf_name, vnf_interface=None, metric='tx_packets'):\r
for connected_sw in self.net.DCNetwork_graph.neighbors(vnf_name):\r
link_dict = self.net.DCNetwork_graph[vnf_name][connected_sw]\r
for link in link_dict:\r
- # logging.info("{0},{1}".format(link_dict[link],vnf_interface))\r
if link_dict[link]['src_port_id'] == vnf_interface:\r
# found the right link and connected switch\r
- # logging.info("{0},{1}".format(link_dict[link]['src_port_id'], vnf_source_interface))\r
- network_metric['mon_port'] = link_dict[link]['dst_port']\r
+ network_metric['mon_port'] = link_dict[link]['dst_port_nr']\r
break\r
\r
if 'mon_port' not in network_metric:\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
\r
logging.info('Started monitoring: {2} on {0}:{1}'.format(vnf_name, vnf_interface, metric))\r
return 'Started monitoring: {2} on {0}:{1}'.format(vnf_name, vnf_interface, metric)\r
logging.exception("setup_metric error.")\r
return ex.message\r
\r
- def stop_metric(self, vnf_name, vnf_interface, metric):\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 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
#this removes the complete metric, all labels...\r
#REGISTRY.unregister(self.prom_metrics[metric_dict['metric_key']])\r
+ #self.registry.unregister(self.prom_metrics[metric_dict['metric_key']])\r
+\r
+ for collector in self.registry._collectors :\r
+\r
+ """\r
+ INFO:root:name:sonemu_rx_count_packets\r
+ labels:('vnf_name', 'vnf_interface')\r
+ metrics:{(u'tsrc', u'output'): < prometheus_client.core.Gauge\r
+ object\r
+ at\r
+ 0x7f353447fd10 >}\r
+ """\r
+ logging.info('{0}'.format(collector._metrics.values()))\r
+\r
+ if (vnf_name, vnf_interface, 'None') in collector._metrics:\r
+ logging.info('2 name:{0} labels:{1} metrics:{2}'.format(collector._name, collector._labelnames,\r
+ collector._metrics))\r
+ collector.remove(vnf_name, vnf_interface, 'None')\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']].labels(vnf_name, vnf_interface).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 them a separate grouping key, and probably a diffferent registry also\r
+ delete_from_gateway(self.pushgateway, job='sonemu-SDNcontroller')\r
+\r
+ self.monitor_lock.release()\r
\r
logging.info('Stopped monitoring: {2} on {0}:{1}'.format(vnf_name, vnf_interface, metric))\r
return 'Stopped monitoring: {2} on {0}:{1}'.format(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
+ for collector in self.registry._collectors:\r
+ collector_dict = collector._metrics.copy()\r
+ for name, interface, id in collector_dict:\r
+ if name == vnf_name:\r
+ logging.info('3 name:{0} labels:{1} metrics:{2}'.format(collector._name, collector._labelnames,\r
+ collector._metrics))\r
+ collector.remove(name, interface, 'None')\r
+\r
+ delete_from_gateway(self.pushgateway, job='sonemu-SDNcontroller')\r
+ self.monitor_lock.release()\r
+ logging.info('Stopped monitoring vnf: {0}'.format(vnf_name))\r
+ return 'Stopped monitoring: {0}'.format(vnf_name)\r
+\r
+ return 'Error stopping monitoring metric: {0} on {1}:{2}'.format(metric, vnf_name, vnf_interface)\r
+\r
+\r
+# get all metrics defined in the list and export it to Prometheus\r
+ def get_flow_metrics(self):\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
+\r
+ # query Ryu\r
+ ret = self.net.ryu_REST('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
+ self.monitor_flow_lock.release()\r
+ time.sleep(1)\r
\r
- # get all metrics defined in the list and export it to Prometheus\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'] for metric_dict in self.network_metrics]\r
dpid_set = set(dpid_list)\r
for dpid in dpid_set:\r
\r
# query Ryu\r
- ret = self.REST_cmd('stats/port', dpid)\r
- port_stat_dict = ast.literal_eval(ret)\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
- #logging.info('1set prom packets:{0} '.format(self.network_metrics))\r
+\r
for metric_dict in metric_list:\r
self.set_network_metric(metric_dict, port_stat_dict)\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 reply\r
if int(port_stat['port_no']) == int(mon_port):\r
port_uptime = port_stat['duration_sec'] + port_stat['duration_nsec'] * 10 ** (-9)\r
this_measurement = int(port_stat[metric_key])\r
- #logging.info('set prom packets:{0} {1}:{2}'.format(this_measurement, vnf_name, vnf_interface))\r
\r
# set prometheus metric\r
- self.prom_metrics[metric_dict['metric_key']].labels(vnf_name, vnf_interface).set(this_measurement)\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
+ # 1 single monitor job for all metrics of the SDN controller\r
+ pushadd_to_gateway(self.pushgateway, job='sonemu-SDNcontroller', registry=self.registry)\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(port_stat[metric_key])\r
metric_dict['previous_monitor_time'] = port_uptime\r
# do first measurement\r
- #logging.info('first measurement')\r
- time.sleep(1)\r
- byte_rate = self.get_network_metrics()\r
- return byte_rate\r
+ #time.sleep(1)\r
+ #self.monitor_lock.release()\r
+ # rate cannot be calculated yet (need a first measurement)\r
+ metric_rate = None\r
+\r
else:\r
time_delta = (port_uptime - metric_dict['previous_monitor_time'])\r
- byte_rate = (this_measurement - metric_dict['previous_measurement']) / float(time_delta)\r
- # logging.info('uptime:{2} delta:{0} rate:{1}'.format(time_delta,byte_rate,port_uptime))\r
+ metric_rate = (this_measurement - metric_dict['previous_measurement']) / float(time_delta)\r
\r
metric_dict['previous_measurement'] = this_measurement\r
metric_dict['previous_monitor_time'] = port_uptime\r
- return byte_rate\r
+ return\r
\r
logging.exception('metric {0} not found on {1}:{2}'.format(metric_key, vnf_name, vnf_interface))\r
+ logging.exception('monport:{0}, dpid:{1}'.format(mon_port, switch_dpid))\r
+ logging.exception('port dict:{0}'.format(port_stat_dict))\r
return 'metric {0} not found on {1}:{2}'.format(metric_key, vnf_name, vnf_interface)\r
\r
+ def set_flow_metric(self, metric_dict, flow_stat_dict):\r
+ # vnf tx is the datacenter switch rx and vice-versa\r
+ metric_key = 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_measurement = metric_dict['previous_measurement']\r
+ previous_monitor_time = metric_dict['previous_monitor_time']\r
+ cookie = metric_dict['cookie']\r
+\r
+ counter = 0\r
+ for flow_stat in flow_stat_dict[str(switch_dpid)]:\r
+ if 'bytes' in metric_key:\r
+ counter += flow_stat['byte_count']\r
+ elif 'packet' in metric_key:\r
+ counter += flow_stat['packet_count']\r
+\r
+ # flow_uptime disabled for now (can give error)\r
+ #flow_stat = flow_stat_dict[str(switch_dpid)][0]\r
+ #flow_uptime = flow_stat['duration_sec'] + flow_stat['duration_nsec'] * 10 ** (-9)\r
+\r
+ self.prom_metrics[metric_dict['metric_key']]. \\r
+ labels({'vnf_name': vnf_name, 'vnf_interface': vnf_interface, 'flow_id': cookie}). \\r
+ set(counter)\r
+ try:\r
+ pushadd_to_gateway(self.pushgateway, job='sonemu-SDNcontroller', registry=self.registry)\r
+ except Exception, e:\r
+ logging.warning("Pushgateway not reachable: {0} {1}".format(Exception, e))\r
\r
- def REST_cmd(self, prefix, dpid):\r
- url = self.REST_api + '/' + str(prefix) + '/' + str(dpid)\r
- req = urllib2.Request(url)\r
- ret = urllib2.urlopen(req).read()\r
- return ret\r
\r
def start_Prometheus(self, port=9090):\r
# prometheus.yml configuration file is located in the same directory as this file\r
"--rm",\r
"-p", "{0}:9090".format(port),\r
"-v", "{0}/prometheus.yml:/etc/prometheus/prometheus.yml".format(os.path.dirname(os.path.abspath(__file__))),\r
+ "-v", "{0}/profile.rules:/etc/prometheus/profile.rules".format(os.path.dirname(os.path.abspath(__file__))),\r
"--name", "prometheus",\r
"prom/prometheus"\r
]\r
logging.info('Start Prometheus container {0}'.format(cmd))\r
return Popen(cmd)\r
\r
- def start_cadvisor(self, port=8090):\r
+ def start_PushGateway(self, port=PUSHGATEWAY_PORT):\r
+ cmd = ["docker",\r
+ "run",\r
+ "-d",\r
+ "-p", "{0}:9091".format(port),\r
+ "--name", "pushgateway",\r
+ "--label", 'com.containernet=""',\r
+ "prom/pushgateway"\r
+ ]\r
+\r
+ logging.info('Start Prometheus Push Gateway container {0}'.format(cmd))\r
+ return Popen(cmd)\r
+\r
+ def start_cAdvisor(self, port=CADVISOR_PORT):\r
cmd = ["docker",\r
"run",\r
"--rm",\r
"--volume=/var/lib/docker/:/var/lib/docker:ro",\r
"--publish={0}:8080".format(port),\r
"--name=cadvisor",\r
+ "--label",'com.containernet=""',\r
"google/cadvisor:latest"\r
]\r
logging.info('Start cAdvisor container {0}'.format(cmd))\r
# stop the monitoring thread\r
self.start_monitoring = False\r
self.monitor_thread.join()\r
+ self.monitor_flow_thread.join()\r
\r
+ # these containers are used for monitoring but are started now outside of son-emu\r
+ '''\r
if self.prometheus_process is not None:\r
logging.info('stopping prometheus container')\r
self.prometheus_process.terminate()\r
self.prometheus_process.kill()\r
self._stop_container('prometheus')\r
+ '''\r
+ if self.pushgateway_process is not None:\r
+ logging.info('stopping pushgateway container')\r
+ #self.pushgateway_process.terminate()\r
+ #self.pushgateway_process.kill()\r
+ self._stop_container('pushgateway')\r
\r
if self.cadvisor_process is not None:\r
logging.info('stopping cadvisor container')\r
- self.cadvisor_process.terminate()\r
- self.cadvisor_process.kill()\r
+ #self.cadvisor_process.terminate()\r
+ #self.cadvisor_process.kill()\r
self._stop_container('cadvisor')\r
\r
def switch_tx_rx(self,metric=''):\r
return metric\r
\r
def _stop_container(self, name):\r
- cmd = ["docker",\r
- "stop",\r
- name]\r
- Popen(cmd).wait()\r
\r
cmd = ["docker",\r
"rm",\r
+ "-f",\r
name]\r
Popen(cmd).wait()\r
\r
-\r