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\r
+from subprocess import Popen, check_call\r
import os\r
-\r
+import docker\r
+import json\r
\r
logging.basicConfig(level=logging.INFO)\r
\r
class DCNetworkMonitor():\r
def __init__(self, net):\r
self.net = net\r
+ self.dockercli = docker.from_env()\r
\r
# pushgateway address\r
self.pushgateway = 'localhost:{0}'.format(PUSHGATEWAY_PORT)\r
self.monitor_flow_lock = threading.Lock()\r
self.network_metrics = []\r
self.flow_metrics = []\r
+ self.skewmon_metrics = {}\r
\r
# start monitoring thread\r
self.start_monitoring = True\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
+ # set metric to NaN\r
+ self.prom_metrics[flow_dict['metric_key']]. \\r
+ labels(vnf_name=vnf_name, vnf_interface=vnf_interface, flow_id=cookie). \\r
+ set(float('nan'))\r
\r
delete_from_gateway(self.pushgateway, job='sonemu-SDNcontroller')\r
\r
network_metric = {}\r
\r
# check if port is specified (vnf:port)\r
- if vnf_interface is None:\r
+ if vnf_interface is None or vnf_interface == '':\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
\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']]. \\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
\r
self.set_flow_metric(flow_dict, flow_stat_dict)\r
\r
+\r
+ try:\r
+ if len(self.flow_metrics) > 0:\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
self.monitor_flow_lock.release()\r
time.sleep(1)\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(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
self.monitor_lock.release()\r
time.sleep(1)\r
\r
previous_measurement = metric_dict['previous_measurement']\r
previous_monitor_time = metric_dict['previous_monitor_time']\r
mon_port = metric_dict['mon_port']\r
-\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'] + 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
+ 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
#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
+ 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
\r
def start_Prometheus(self, port=9090):\r
# prometheus.yml configuration file is located in the same directory as this file\r
"--publish={0}:8080".format(port),\r
"--name=cadvisor",\r
"--label",'com.containernet=""',\r
- "google/cadvisor:latest"\r
+ "google/cadvisor:latest",\r
+ "--storage_duration=1m0s",\r
+ "--allow_dynamic_housekeeping=true",\r
+ #"--housekeeping_interval=1s",\r
]\r
logging.info('Start cAdvisor container {0}'.format(cmd))\r
return Popen(cmd)\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
+\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._stop_container('cadvisor')\r
\r
def switch_tx_rx(self,metric=''):\r
\r
def _stop_container(self, name):\r
\r
- cmd = ["docker",\r
- "rm",\r
- "-f",\r
- name]\r
- Popen(cmd).wait()\r
+ #container = self.dockercli.containers.get(name)\r
+ #container.stop()\r
+ #container.remove(force=True)\r
+\r
+ # the only robust way to stop these containers is via Popen, it seems\r
+ time.sleep(1)\r
+ cmd = ['docker', 'rm', '-f', name]\r
+ Popen(cmd)\r
+\r
+\r
+ def update_skewmon(self, vnf_name, resource_name, action):\r
+\r
+ ret = ''\r
+\r
+ config_file_path = '/tmp/skewmon.cfg'\r
+ configfile = open(config_file_path, 'a+')\r
+ try:\r
+ config = json.load(configfile)\r
+ except:\r
+ #not a valid json file or empty\r
+ config = {}\r
+\r
+ #initialize config file\r
+ if len(self.skewmon_metrics) == 0:\r
+ config = {}\r
+ json.dump(config, configfile)\r
+ configfile.close()\r
+\r
+ docker_name = 'mn.' + vnf_name\r
+ vnf_container = self.dockercli.containers.get(docker_name)\r
+ key = resource_name + '_' + vnf_container.short_id\r
+ vnf_id = vnf_container.id\r
+\r
+ if action == 'start':\r
+ # add a new vnf to monitor\r
+ config[key] = dict(VNF_NAME=vnf_name,\r
+ VNF_ID=vnf_id,\r
+ VNF_METRIC=resource_name)\r
+ ret = 'adding to skewness monitor: {0} {1} '.format(vnf_name, resource_name)\r
+ logging.info(ret)\r
+ elif action == 'stop':\r
+ # remove vnf to monitor\r
+ config.pop(key)\r
+ ret = 'removing from skewness monitor: {0} {1} '.format(vnf_name, resource_name)\r
+ logging.info(ret)\r
+\r
+ self.skewmon_metrics = config\r
+ configfile = open(config_file_path, 'w')\r
+ json.dump(config, configfile)\r
+ configfile.close()\r
+\r
+ try:\r
+ skewmon_container = self.dockercli.containers.get('skewmon')\r
+\r
+ # remove container if config is empty\r
+ if len(config) == 0:\r
+ ret += 'stopping skewness monitor'\r
+ logging.info('stopping skewness monitor')\r
+ skewmon_container.remove(force=True)\r
+\r
+ except docker.errors.NotFound:\r
+ # start container if not running\r
+ ret += 'starting skewness monitor'\r
+ logging.info('starting skewness monitor')\r
+ volumes = {'/sys/fs/cgroup':{'bind':'/sys/fs/cgroup', 'mode':'ro'},\r
+ '/tmp/skewmon.cfg':{'bind':'/config.txt', 'mode':'ro'}}\r
+ self.dockercli.containers.run('skewmon',\r
+ detach=True,\r
+ volumes=volumes,\r
+ labels=['com.containernet'],\r
+ name='skewmon'\r
+ )\r
+ # Wait a while for containers to be completely started\r
+ started = False\r
+ wait_time = 0\r
+ while not started:\r
+ list1 = self.dockercli.containers.list(filters={'status': 'running', 'name': 'prometheus'})\r
+ if len(list1) >= 1:\r
+ time.sleep(1)\r
+ started = True\r
+ if wait_time > 5:\r
+ return 'skewmon not started'\r
+ time.sleep(1)\r
+ wait_time += 1\r
+ return ret\r
+\r
+\r
+\r
+\r
\r