# limitations under the License.
""" Resource collection definitions """
-from __future__ import absolute_import
-from __future__ import print_function
-
-import logging
+import errno
from itertools import chain
-
-import jinja2
+import logging
+import multiprocessing
import os
import os.path
import re
-import multiprocessing
-import pkg_resources
+import jinja2
+import pkg_resources
from oslo_config import cfg
from oslo_utils.encodeutils import safe_decode
from yardstick.common.task_template import finalize_for_yaml
from yardstick.common.utils import validate_non_string_sequence
from yardstick.network_services.nfvi.collectd import AmqpConsumer
-from yardstick.network_services.utils import get_nsb_option
+
LOG = logging.getLogger(__name__)
COLLECTD_CONF = "collectd.conf"
AMPQ_PORT = 5672
DEFAULT_INTERVAL = 25
+ DEFAULT_TIMEOUT = 3600
+ OVS_SOCKET_PATH = "/usr/local/var/run/openvswitch/db.sock"
+
+ def __init__(self, mgmt, port_names=None, plugins=None, interval=None, timeout=None):
- def __init__(self, mgmt, port_names=None, cores=None, plugins=None, interval=None):
if plugins is None:
self.plugins = {}
else:
self.plugins = plugins
+
if interval is None:
self.interval = self.DEFAULT_INTERVAL
else:
self.interval = interval
+
+ if timeout is None:
+ self.timeout = self.DEFAULT_TIMEOUT
+ else:
+ self.timeout = timeout
+
self.enable = True
- self.cores = validate_non_string_sequence(cores, default=[])
self._queue = multiprocessing.Queue()
self.amqp_client = None
self.port_names = validate_non_string_sequence(port_names, default=[])
self.mgmt = mgmt
self.connection = ssh.AutoConnectSSH.from_node(mgmt)
- def check_if_sa_running(self, process):
+ @classmethod
+ def make_from_node(cls, node, timeout):
+ # node dict works as mgmt dict
+ # don't need port names, there is no way we can
+ # tell what port is used on the compute node
+ collectd_options = node["collectd"]
+ plugins = collectd_options.get("plugins", {})
+ interval = collectd_options.get("interval")
+
+ return cls(node, plugins=plugins, interval=interval, timeout=timeout)
+
+ def check_if_system_agent_running(self, process):
""" verify if system agent is running """
- err, pid, _ = self.connection.execute("pgrep -f %s" % process)
- return [err == 0, pid]
+ try:
+ err, pid, _ = self.connection.execute("pgrep -f %s" % process)
+ # strip whitespace
+ return err, pid.strip()
+ except OSError as e:
+ if e.errno in {errno.ECONNRESET}:
+ # if we can't connect to check, then we won't be able to connect to stop it
+ LOG.exception("Can't connect to host to check %s status", process)
+ return 1, None
+ raise
def run_collectd_amqp(self):
""" run amqp consumer to collect the NFVi data """
def parse_intel_pmu_stats(cls, key, value):
return {''.join(str(v) for v in key): value.split(":")[1]}
- def parse_collectd_result(self, metrics, core_list):
+ def parse_collectd_result(self, metrics):
""" convert collectd data into json"""
result = {
"cpu": {},
"dpdkstat": {},
"virt": {},
"ovs_stats": {},
- "intel_pmu": {},
}
testcase = ""
res_key0 = next(res_key_iter)
res_key1 = next(res_key_iter)
- if "cpu" in res_key0 or "intel_rdt" in res_key0:
+ if "cpu" in res_key0 or "intel_rdt" in res_key0 or "intel_pmu" in res_key0:
cpu_key, name, metric, testcase = \
self.get_cpu_data(res_key0, res_key1, value)
- if cpu_key in core_list:
- result["cpu"].setdefault(cpu_key, {}).update({name: metric})
+ result["cpu"].setdefault(cpu_key, {}).update({name: metric})
elif "memory" in res_key0:
result["memory"].update({res_key1: value.split(":")[0]})
elif "ovs_stats" in res_key0:
result["ovs_stats"].update(self.parse_ovs_stats(key_split, value))
- elif "intel_pmu-all" in res_key0:
- result["intel_pmu"].update(self.parse_intel_pmu_stats(res_key1, value))
-
result["timestamp"] = testcase
return result
def amqp_process_for_nfvi_kpi(self):
""" amqp collect and return nfvi kpis """
if self.amqp_client is None and self.enable:
- self.amqp_client = \
- multiprocessing.Process(target=self.run_collectd_amqp)
+ self.amqp_client = multiprocessing.Process(
+ name="AmqpClient-{}-{}".format(self.mgmt['ip'], os.getpid()),
+ target=self.run_collectd_amqp)
self.amqp_client.start()
def amqp_collect_nfvi_kpi(self):
metric = {}
while not self._queue.empty():
metric.update(self._queue.get())
- msg = self.parse_collectd_result(metric, self.cores)
+ msg = self.parse_collectd_result(metric)
return msg
def _provide_config_file(self, config_file_path, nfvi_cfg, template_kwargs):
}
self._provide_config_file(config_file_path, self.COLLECTD_CONF, kwargs)
+ def _setup_ovs_stats(self, connection):
+ try:
+ socket_path = self.plugins["ovs_stats"].get("ovs_socket_path", self.OVS_SOCKET_PATH)
+ except KeyError:
+ # ovs_stats is not a dict
+ socket_path = self.OVS_SOCKET_PATH
+ status = connection.execute("test -S {}".format(socket_path))[0]
+ if status != 0:
+ LOG.error("cannot find OVS socket %s", socket_path)
+
def _start_collectd(self, connection, bin_path):
LOG.debug("Starting collectd to collect NFVi stats")
connection.execute('sudo pkill -x -9 collectd')
- bin_path = get_nsb_option("bin_path")
collectd_path = os.path.join(bin_path, "collectd", "sbin", "collectd")
config_file_path = os.path.join(bin_path, "collectd", "etc")
exit_status = connection.execute("which %s > /dev/null 2>&1" % collectd_path)[0]
# connection.execute("sudo %s '%s' '%s'" % (
# collectd_installer, http_proxy, https_proxy))
return
+ if "ovs_stats" in self.plugins:
+ self._setup_ovs_stats(connection)
+
LOG.debug("Starting collectd to collect NFVi stats")
# ensure collectd.conf.d exists to avoid error/warning
connection.execute("sudo mkdir -p /etc/collectd/collectd.conf.d")
connection.execute("sudo rabbitmqctl authenticate_user admin admin")
connection.execute("sudo rabbitmqctl set_permissions -p / admin '.*' '.*' '.*'")
- LOG.debug("Start collectd service.....")
- connection.execute("sudo %s" % collectd_path)
+ LOG.debug("Start collectd service..... %s second timeout", self.timeout)
+ # intel_pmu plug requires large numbers of files open, so try to set
+ # ulimit -n to a large value
+ connection.execute("sudo bash -c 'ulimit -n 1000000 ; %s'" % collectd_path,
+ timeout=self.timeout)
LOG.debug("Done")
def initiate_systemagent(self, bin_path):
LOG.debug("Stop resource monitor...")
if self.amqp_client is not None:
+ # we proper and try to join first
+ self.amqp_client.join(3)
self.amqp_client.terminate()
- status, pid = self.check_if_sa_running(agent)
- if status == 0:
+ LOG.debug("Check if %s is running", agent)
+ status, pid = self.check_if_system_agent_running(agent)
+ LOG.debug("status %s pid %s", status, pid)
+ if status != 0:
return
- self.connection.execute('sudo kill -9 %s' % pid)
- self.connection.execute('sudo pkill -9 %s' % agent)
+ if pid:
+ self.connection.execute('sudo kill -9 "%s"' % pid)
+ self.connection.execute('sudo pkill -9 "%s"' % agent)
self.connection.execute('sudo service rabbitmq-server stop')
self.connection.execute("sudo rabbitmqctl stop_app")