start cadvisor and prometheus docker container at startup
diff --git a/src/emuvim/api/zerorpc/network.py b/src/emuvim/api/zerorpc/network.py
index 1b7e1e0..f21ce99 100644
--- a/src/emuvim/api/zerorpc/network.py
+++ b/src/emuvim/api/zerorpc/network.py
@@ -100,6 +100,16 @@
logging.exception("RPC error.")
return ex.message
+ # remove the rate measurement for a vnf interface
+ def remove_metric(self, vnf_name, vnf_interface, metric):
+ logging.debug("RPC CALL: setup metric")
+ try:
+ c = self.net.monitor_agent.remove_metric(vnf_name, vnf_interface, metric)
+ return c
+ except Exception as ex:
+ logging.exception("RPC error.")
+ return ex.message
+
# setup the rate measurement for a vnf interface
def monitor_setup_rate_measurement(self, vnf_name, vnf_interface, metric):
logging.debug("RPC CALL: get rate")
diff --git a/src/emuvim/cli/monitor.py b/src/emuvim/cli/monitor.py
index 1d4abfb..bae0c07 100755
--- a/src/emuvim/cli/monitor.py
+++ b/src/emuvim/cli/monitor.py
@@ -36,19 +36,15 @@
vnf_interface,
args.get("metric"))
pp.pprint(r)
- '''
- self.c.monitor_setup_rate_measurement(
+
+ def remove_metric(self, args):
+ vnf_name = self._parse_vnf_name(args.get("vnf_name"))
+ vnf_interface = self._parse_vnf_interface(args.get("vnf_name"))
+ r = self.c.remove_metric(
vnf_name,
vnf_interface,
args.get("metric"))
- while True:
- r = self.c.monitor_get_rate(
- vnf_name,
- vnf_interface,
- args.get("metric"))
- pp.pprint(r)
- time.sleep(1)
- '''
+ pp.pprint(r)
def _parse_vnf_name(self, vnf_name_str):
vnf_name = vnf_name_str.split(':')[0]
diff --git a/src/emuvim/dcemulator/monitoring.py b/src/emuvim/dcemulator/monitoring.py
index d745069..7fa5e18 100755
--- a/src/emuvim/dcemulator/monitoring.py
+++ b/src/emuvim/dcemulator/monitoring.py
@@ -7,8 +7,8 @@
import time
from prometheus_client import start_http_server, Summary, Histogram, Gauge, Counter
import threading
-from subprocess import Popen
-from os import getcwd
+from subprocess import Popen, PIPE
+import os
logging.basicConfig(level=logging.INFO)
@@ -25,16 +25,6 @@
self.REST_api = 'http://{0}:{1}'.format(self.ip,self.port)
# helper variables to calculate the metrics
- # TODO put these in a list to support multiple metrics simultaneously
- self.switch_dpid = 0
- self.vnf_name = None
- self.vnf_interface = None
- self.previous_measurement = 0
- self.previous_monitor_time = 0
- self.metric_key = None
- self.mon_port = None
-
-
# Start up the server to expose the metrics to Prometheus.
start_http_server(8000)
# supported Prometheus metrics
@@ -66,13 +56,13 @@
self.network_metrics = []
# start monitoring thread
+ self.start_monitoring = True
self.monitor_thread = threading.Thread(target=self.get_network_metrics)
self.monitor_thread.start()
# helper tools
- self.prometheus_process = None
- self.cAdvisor_process = None
-
+ self.prometheus_process = self.start_Prometheus()
+ self.cadvisor_process = self.start_cadvisor()
# first set some parameters, before measurement can start
def setup_metric(self, vnf_name, vnf_interface=None, metric='tx_packets'):
@@ -88,8 +78,6 @@
network_metric['vnf_name'] = vnf_name
network_metric['vnf_interface'] = vnf_interface
- #self.vnf_name = vnf_name
- #self.vnf_interface = vnf_interface
for connected_sw in self.net.DCNetwork_graph.neighbors(vnf_name):
link_dict = self.net.DCNetwork_graph[vnf_name][connected_sw]
@@ -99,7 +87,6 @@
# found the right link and connected switch
# logging.info("{0},{1}".format(link_dict[link]['src_port_id'], vnf_source_interface))
network_metric['mon_port'] = link_dict[link]['dst_port']
- # self.mon_port = link_dict[link]['dst_port']
break
if 'mon_port' not in network_metric:
@@ -129,13 +116,11 @@
network_metric['previous_measurement'] = 0
network_metric['previous_monitor_time'] = 0
- #self.previous_measurement = 0
- #self.previous_monitor_time = 0
+
network_metric['switch_dpid'] = int(str(next_node.dpid), 16)
network_metric['metric_key'] = metric
- #self.switch_dpid = int(str(next_node.dpid), 16)
- #self.metric_key = '{0}_{1}'.format(direction, metric)
+
self.network_metrics.append(network_metric)
@@ -146,10 +131,18 @@
logging.exception("get_rate error.")
return ex.message
+ def remove_metric(self, vnf_name, vnf_interface, metric):
+ for metric_dict in self.network_metrics:
+ if metric_dict['vnf_name'] == vnf_name and metric_dict['vnf_interface'] == vnf_interface \
+ and metric_dict['metric'] == metric:
+ self.network_metrics.remove(metric_dict)
+ logging.info('Stopped monitoring: {2} on {0}:{1}'.format(vnf_name, vnf_interface, metric))
+ return 'Stopped monitoring: {2} on {0}:{1}'.format(vnf_name, vnf_interface, metric)
- # get all metrics defined in the list
+
+ # get all metrics defined in the list and export it to Prometheus
def get_network_metrics(self):
- while True:
+ while self.start_monitoring:
# group metrics by dpid to optimize the rest api calls
dpid_list = [metric_dict['switch_dpid'] for metric_dict in self.network_metrics]
dpid_set = set(dpid_list)
@@ -168,10 +161,10 @@
time.sleep(1)
- # call this function repeatedly for streaming measurements
+ # add metric to the list to export to Prometheus, parse the Ryu port-stats reply
def set_network_metric(self, metric_dict, port_stat_dict):
-
- metric_key = metric_dict['metric_key']
+ # vnf tx is the datacenter switch rx and vice-versa
+ metric_key = self.switch_tx_rx(metric_dict['metric_key'])
switch_dpid = metric_dict['switch_dpid']
vnf_name = metric_dict['vnf_name']
vnf_interface = metric_dict['vnf_interface']
@@ -209,43 +202,6 @@
return 'metric {0} not found on {1}:{2}'.format(metric_key, vnf_name, vnf_interface)
- # call this function repeatedly for streaming measurements
- def get_rate(self, vnf_name, vnf_interface=None, direction='tx', metric='packets'):
-
- key = self.metric_key
-
- ret = self.REST_cmd('stats/port', self.switch_dpid)
- port_stat_dict = ast.literal_eval(ret)
- for port_stat in port_stat_dict[str(self.switch_dpid)]:
- if port_stat['port_no'] == self.mon_port:
- port_uptime = port_stat['duration_sec'] + port_stat['duration_nsec'] * 10 ** (-9)
- this_measurement = int(port_stat[key])
- #logging.info('packets:{0}'.format(this_measurement))
-
- # set prometheus metrics
- if metric == 'packets':
- self.prom_tx_packet_count.labels(self.vnf_name, self.vnf_interface).set(this_measurement)
- elif metric == 'bytes':
- self.prom_tx_byte_count.labels(self.vnf_name, self.vnf_interface).set(this_measurement)
-
- if self.previous_monitor_time <= 0 or self.previous_monitor_time >= port_uptime:
- self.previous_measurement = int(port_stat[key])
- self.previous_monitor_time = port_uptime
- # do first measurement
- time.sleep(1)
- byte_rate = self.get_rate(vnf_name, vnf_interface, direction, metric)
- return byte_rate
- else:
- time_delta = (port_uptime - self.previous_monitor_time)
- byte_rate = (this_measurement - self.previous_measurement) / float(time_delta)
- #logging.info('uptime:{2} delta:{0} rate:{1}'.format(time_delta,byte_rate,port_uptime))
-
- self.previous_measurement = this_measurement
- self.previous_monitor_time = port_uptime
- return byte_rate
-
- return ret
-
def REST_cmd(self, prefix, dpid):
url = self.REST_api + '/' + str(prefix) + '/' + str(dpid)
req = urllib2.Request(url)
@@ -253,18 +209,19 @@
return ret
def start_Prometheus(self, port=9090):
+ # prometheus.yml configuration file is located in the same directory as this file
cmd = ["docker",
"run",
"--rm",
"-p", "{0}:9090".format(port),
- "-v", "{0}/prometheus.yml:/etc/prometheus/prometheus.yml".format(getcwd()),
+ "-v", "{0}/prometheus.yml:/etc/prometheus/prometheus.yml".format(os.path.dirname(os.path.abspath(__file__))),
"--name", "prometheus",
"prom/prometheus"
]
+ logging.info('Start Prometheus container {0}'.format(cmd))
+ return Popen(cmd)
- self.prometheus_process = Popen(cmd)
-
- def start_cAdvisor(self, port=8090):
+ def start_cadvisor(self, port=8090):
cmd = ["docker",
"run",
"--rm",
@@ -276,13 +233,45 @@
"--name=cadvisor",
"google/cadvisor:latest"
]
- self.cAdvisor_process = Popen(cmd)
+ logging.info('Start cAdvisor container {0}'.format(cmd))
+ return Popen(cmd)
def stop(self):
+ # stop the monitoring thread
+ self.start_monitoring = False
+ self.monitor_thread.join()
+
if self.prometheus_process is not None:
+ logging.info('stopping prometheus container')
self.prometheus_process.terminate()
self.prometheus_process.kill()
+ self._stop_container('prometheus')
- if self.cAdvisor_process is not None:
- self.cAdvisor_process.terminate()
- self.cAdvisor_process.kill()
+ if self.cadvisor_process is not None:
+ logging.info('stopping cadvisor container')
+ self.cadvisor_process.terminate()
+ self.cadvisor_process.kill()
+ self._stop_container('cadvisor')
+
+ def switch_tx_rx(self,metric=''):
+ # when monitoring vnfs, the tx of the datacenter switch is actually the rx of the vnf
+ # so we need to change the metric name to be consistent with the vnf rx or tx
+ if 'tx' in metric:
+ metric = metric.replace('tx','rx')
+ elif 'rx' in metric:
+ metric = metric.replace('rx','tx')
+
+ return metric
+
+ def _stop_container(self, name):
+ cmd = ["docker",
+ "stop",
+ name]
+ Popen(cmd).wait()
+
+ cmd = ["docker",
+ "rm",
+ name]
+ Popen(cmd).wait()
+
+