--- /dev/null
+.. This work is licensed under a Creative Commons Attribution 4.0 International License.
+.. http://creativecommons.org/licenses/by/4.0
+
+Monitor Types and Limitations
+=============================
+
+Currently there are two monitor types supported: sample and collectd
+
+Sample Monitor
+--------------
+
+Sample monitor type pings the compute host from the control host and calculates the
+notification time after the ping timeout.
+Also if inspector type is sample, the compute node needs to communicate with the control
+node on port 12345. This port needs to be opened for incomming traffic on control node.
+
+Collectd Monitor
+----------------
+
+Collectd monitor type uses collectd daemon running ovs_events plugin. Collectd runs on
+compute to send instant notification to the control node. The notification time is
+calculated by using the difference of time at which compute node sends notification to
+control node and the time at which consumer is notified. The time on control and compute
+node has to be synchronized for this reason. For further details on setting up collectd
+on the compute node, use the following link:
+http://docs.opnfv.org/en/stable-danube/submodules/barometer/docs/release/userguide/feature.userguide.html#id18
+
+Collectd monitors an interface managed by OVS. If the interface is not be assigned
+an IP, the user has to provide the name of interface to be monitored. The command to
+launch the doctor test in that case is:
+MONITOR_TYPE=collectd INSPECTOR_TYPE=sample INTERFACE_NAME=example_iface ./run.sh
+
+If the interface name or IP is not provided, the collectd monitor type will monitor the
+default management interface. This may result in the failure of doctor run.sh test case.
+The test case sets the monitored interface down and if the inspector (sample or congress)
+is running on the same subnet, collectd monitor will not be able to communicate with it.
quota_instances = 1
quota_cores = 1
-
+[monitor]
+type=collectd
def main():
args = get_args()
- app.run(port=args.port)
-
+ app.run(host='0.0.0.0', port=args.port)
if __name__ == '__main__':
main()
--- /dev/null
+#!/bin/bash
+
+MONITOR_TYPE=${MONITOR_TYPE:-sample}
+
+function is_monitor_supported {
+ local monitor="$1"
+ [[ -f $TOP_DIR/lib/monitors/$monitor/$monitor ]]
+}
+
+function is_monitor {
+ local monitor="$1"
+ [[ $monitor == $MONITOR_TYPE ]]
+}
+
+function start_monitor {
+ start_monitor_$MONITOR_TYPE
+}
+
+function stop_monitor {
+ stop_monitor_$MONITOR_TYPE
+}
+
+function cleanup_monitor {
+ cleanup_monitor_$MONITOR_TYPE
+}
+
+if ! is_monitor_supported $MONITOR_TYPE; then
+ die $LINENO "MONITOR_TYPE=$MONITOR_TYPE is not supported."
+fi
+
+source $TOP_DIR/lib/monitors/$MONITOR_TYPE/$MONITOR_TYPE
--- /dev/null
+#!/bin/bash
+
+function start_monitor_collectd {
+ ## CONTROL_IP is the IP of primary interface of control node i.e.
+ ## eth0, eno1. It is used by collectd monitor to communicate with
+ ## sample inspector.
+ ## @TODO (umar) see if mgmt IP of control is a better option. Also
+ ## primary interface may not be the right option
+ CONTROL_IP="$(ip a | sed -En 's/127.0.0.1//;s/.*inet (addr:)?(([0-9]*\.){3}[0-9]*).*/\2/p' | sed -n 1p)"
+ #CONTROL_IP=192.168.98.6
+
+ echo "
+Hostname \"$COMPUTE_HOST\"
+FQDNLookup false
+Interval 1
+MaxReadInterval 2
+
+<LoadPlugin python>
+ Globals true
+</LoadPlugin>
+LoadPlugin ovs_events
+LoadPlugin logfile
+
+<Plugin logfile>
+ File \"/var/log/collectd.log\"
+ Timestamp true
+ LogLevel \"info\"
+</Plugin>
+
+<Plugin python>
+ ModulePath \"/home/$COMPUTE_USER\"
+ LogTraces true
+ Interactive false
+ Import \"collectd_plugin\"
+ <Module \"collectd_plugin\">
+ control_ip \"$CONTROL_IP\"
+ compute_ip \"$COMPUTE_IP\"
+ compute_host \"$COMPUTE_HOST\"
+ compute_user \"$COMPUTE_USER\"
+ inspector_type \"$INSPECTOR_TYPE\"
+ os_auth_url \"$OS_AUTH_URL\"
+ os_username \"$OS_USERNAME\"
+ os_password \"$OS_PASSWORD\"
+ os_project_name \"$OS_PROJECT_NAME\"
+ os_user_domain_name \"$OS_USER_DOMAIN_NAME\"
+ os_user_domain_id \"$OS_USER_DOMAIN_ID\"
+ os_project_domain_name \"$OS_PROJECT_DOMAIN_NAME\"
+ os_project_domain_id \"$OS_PROJECT_DOMAIN_ID\"
+ </Module>
+</Plugin>
+
+<Plugin ovs_events>
+ Port 6640
+ Socket \"/var/run/openvswitch/db.sock\"
+ Interfaces \"@INTERFACE_NAME@\"
+ SendNotification true
+ DispatchValues false
+</Plugin>
+
+" > $TOP_DIR/lib/monitors/collectd.conf
+
+ scp $ssh_opts_cpu $TOP_DIR/lib/monitors/collectd.conf $COMPUTE_USER@$COMPUTE_IP:
+ ## @TODO (umar) Always assuming that the interface is assigned an IP if
+ ## interface name is not provided. See if there is a better approach
+ ssh $ssh_opts_cpu "$COMPUTE_USER@$COMPUTE_IP" "
+ if [ -n \"$INTERFACE_NAME\" ]; then
+ dev=$INTERFACE_NAME
+ else
+ dev=\$(sudo ip a | awk '/ $COMPUTE_IP\//{print \$NF}')
+ fi
+ sed -i -e \"s/@INTERFACE_NAME@/\$dev/\" collectd.conf
+ collectd_conf=/opt/collectd/etc/collectd.conf
+ if [ -e \$collectd_conf ]; then
+ sudo cp \$collectd_conf \${collectd_conf}-doctor-saved
+ else
+ sudo touch \${collectd_conf}-doctor-created
+ fi
+ sudo mv collectd.conf /opt/collectd/etc/collectd.conf"
+
+ scp $ssh_opts_cpu $TOP_DIR/lib/monitors/collectd/collectd_plugin.py $COMPUTE_USER@$COMPUTE_IP:collectd_plugin.py
+ ssh $ssh_opts_cpu "$COMPUTE_USER@$COMPUTE_IP" "sudo pkill collectd
+ sudo /opt/collectd/sbin/collectd"
+}
+
+function stop_monitor_collectd {
+ ssh $ssh_opts_cpu "$COMPUTE_USER@$COMPUTE_IP" 'sudo pkill collectd'
+}
+
+function cleanup_monitor_collectd {
+ ssh $ssh_opts_cpu "$COMPUTE_USER@$COMPUTE_IP" "
+ collectd_conf=/opt/collectd/etc/collectd.conf
+ if [ -e \"\${collectd_conf}-doctor-created\" ]; then
+ sudo rm \"\${collectd_conf}-doctor-created\"
+ sudo rm \$collectd_conf
+ elif [ -e \"\${collectd_conf}-doctor-saved\" ]; then
+ sudo cp -f \"\${collectd_conf}-doctor-saved\" \$collectd_conf
+ sudo rm \"\${collectd_conf}-doctor-saved\"
+ fi"
+
+ rm $TOP_DIR/lib/monitors/collectd.conf
+}
--- /dev/null
+##############################################################################
+# Copyright (c) 2017 NEC Corporation and others.
+#
+# All rights reserved. This program and the accompanying materials
+# are made available under the terms of the Apache License, Version 2.0
+# which accompanies this distribution, and is available at
+# http://www.apache.org/licenses/LICENSE-2.0
+##############################################################################
+
+import collectd
+import sys
+from netifaces import interfaces, ifaddresses, AF_INET
+from datetime import datetime
+import json
+import requests
+import time
+from requests.exceptions import ConnectionError
+
+from keystoneauth1 import loading
+from keystoneauth1 import session
+from congressclient.v1 import client
+
+
+def write_debug(str_write, write_type, compute_user):
+ file_name = ('/home/%s/monitor.log' % compute_user)
+ file_tmp = open(file_name, write_type)
+ file_tmp.write( "%s" % str_write)
+ file_tmp.close()
+
+
+class DoctorMonitorCollectd(object):
+ def __init__(self):
+ self.control_ip = ''
+ self.compute_user = ''
+ self.compute_ip = ''
+ self.host_name = ''
+ self.inspector_type = ''
+ self.inspector_url = ''
+ self.os_auth_url = ''
+ self.os_username = ''
+ self.os_password = ''
+ self.os_project_name = ''
+ self.os_user_domain_name = ''
+ self.os_user_domain_id = ''
+ self.os_project_domain_name = ''
+ self.os_project_domain_id = ''
+ self.sess = ''
+ self.auth = ''
+ self.inspector_notified = 0
+ self.start_notifications = 0
+ self.monitor_type = 'sample'
+
+ def config_func(self, config):
+ for node in config.children:
+ key = node.key.lower()
+ val = node.values[0]
+
+ if key == 'compute_host':
+ self.host_name = val
+ elif key == 'control_ip':
+ self.control_ip = val
+ elif key == 'compute_ip':
+ self.compute_ip = val
+ elif key == 'compute_user':
+ self.compute_user = val
+ elif key == 'inspector_type':
+ self.inspector_type = val
+ elif key == 'os_auth_url':
+ self.os_auth_url = val
+ elif key == 'os_username':
+ self.os_username = val
+ elif key == 'os_password':
+ self.os_password = val
+ elif key == 'os_project_name':
+ self.os_project_name = val
+ elif key == 'os_user_domain_name':
+ self.os_user_domain_name = val
+ elif key == 'os_user_domain_id':
+ self.os_user_domain_id = val
+ elif key == 'os_project_domain_name':
+ self.os_project_domain_name = val
+ elif key == 'os_project_domain_id':
+ self.os_project_domain_id = val
+ else:
+ collectd.info('Unknown config key "%s"' % key)
+
+ def init_collectd(self):
+ write_debug("Compute node collectd monitor start at %s\n\n" % datetime.now().isoformat(), "w", self.compute_user)
+
+ if self.inspector_type == 'sample':
+ self.inspector_url = ('http://%s:12345/events' % self.control_ip)
+ elif self.inspector_type == 'congress':
+ loader = loading.get_plugin_loader('password')
+ self.auth = loader.load_from_options(auth_url=self.os_auth_url,
+ username=self.os_username,
+ password=self.os_password,
+ project_name=self.os_project_name,
+ user_domain_name=self.os_user_domain_name,
+ user_domain_id=self.os_user_domain_id,
+ project_domain_name=self.os_project_domain_name,
+ project_domain_id=self.os_project_domain_id)
+ self.sess=session.Session(auth=self.auth)
+ congress = client.Client(session=self.sess, service_type='policy')
+ ds = congress.list_datasources()['results']
+ doctor_ds = next((item for item in ds if item['driver'] == 'doctor'),
+ None)
+
+ congress_endpoint = congress.httpclient.get_endpoint(auth=self.auth)
+ self.inspector_url = ('%s/v1/data-sources/%s/tables/events/rows' %
+ (congress_endpoint, doctor_ds['id']))
+ else:
+ sys.exit()
+ self.start_notifications = 1
+
+
+ def notify_inspector(self):
+ event_type = "compute.host.down"
+ payload = [
+ {
+ 'id': ("monitor_%s_id1" % self.monitor_type),
+ 'time': datetime.now().isoformat(),
+ 'type': event_type,
+ 'details': {
+ 'hostname': self.host_name,
+ 'status': 'down',
+ 'monitor': ("monitor_%s" % self.monitor_type),
+ 'monitor_event_id': ("monitor_%s_event1" % self.monitor_type)
+ },
+ },
+ ]
+ data = json.dumps(payload)
+ self.inspector_notified = 1
+
+ if self.inspector_type == 'sample':
+ headers = {'content-type': 'application/json'}
+ try:
+ requests.post(self.inspector_url, data=data, headers=headers)
+ except ConnectionError as err:
+ print err
+ elif self.inspector_type == 'congress':
+ # TODO(umar) enhance for token expiry case
+ headers = {
+ 'Content-Type': 'application/json',
+ 'Accept': 'application/json',
+ 'X-Auth-Token': self.sess.get_token()
+ }
+ requests.put(self.inspector_url, data=data, headers=headers)
+
+
+ def handle_notif(self, notification, data=None):
+ if (notification.severity == collectd.NOTIF_FAILURE or
+ notification.severity == collectd.NOTIF_WARNING):
+ if (self.start_notifications == 1 and self.inspector_notified == 0):
+ write_debug("Received down notification: doctor monitor detected at %s\n" % time.time(), "a", self.compute_user)
+ self.notify_inspector()
+
+ elif notification.severity == collectd.NOTIF_OKAY:
+ collectd.info("Interface status: UP again %s\n" % time.time())
+ else:
+ collectd.info("Unknown notification severity %s\n" % notification.severity)
+
+
+monitor = DoctorMonitorCollectd()
+
+collectd.register_config(monitor.config_func)
+collectd.register_init(monitor.init_collectd)
+collectd.register_notification(monitor.handle_notif)
--- /dev/null
+#!/bin/bash
+
+function start_monitor_sample {
+ cp $TOP_DIR/lib/monitors/sample/monitor.py $TOP_DIR/monitor.py
+ pgrep -f "python monitor.py" && return 0
+ sudo -E python monitor.py "$COMPUTE_HOST" "$COMPUTE_IP" "$INSPECTOR_TYPE" \
+ > monitor.log 2>&1 &
+}
+
+function stop_monitor_sample {
+ pgrep -f "python monitor.py" || return 0
+ sudo kill $(pgrep -f "python monitor.py")
+}
+
+function cleanup_monitor_sample {
+ rm monitor.py
+ return
+}
self.setup()
# injecting host failure...
+ # NOTE (umar) add INTERFACE_NAME logic to host injection
# verify the test results
+ # NOTE (umar) copy remote monitor.log file when monitor=collectd
except Exception as e:
LOG.error('doctor test failed, Exception=%s' % e)
OPTS = [
cfg.StrOpt('type',
default='sample',
- choices=['sample'],
+ choices=['sample', 'collectd'],
help='the type of doctor monitor component',
required=True),
]
_monitor_name_class_mapping = {
- 'sample': 'monitor.sample.SampleMonitor'
+ 'sample': 'monitor.sample.SampleMonitor',
+ 'collectd': 'monitor.collectd.CollectdMonitor'
}
def get_monitor(conf, inspector_url, log):
--- /dev/null
+##############################################################################
+# Copyright (c) 2017 NEC Corporation and others.
+#
+# All rights reserved. This program and the accompanying materials
+# are made available under the terms of the Apache License, Version 2.0
+# which accompanies this distribution, and is available at
+# http://www.apache.org/licenses/LICENSE-2.0
+##############################################################################
+
+import os
+import socket
+import getpass
+import sys
+
+from identity_auth import get_session
+from os_clients import nova_client
+from monitor.base import BaseMonitor
+
+
+class CollectdMonitor(BaseMonitor):
+ def __init__(self, conf, inspector_url, log):
+ super(CollectdMonitor, self).__init__(conf, inspector_url, log)
+ self.top_dir = os.path.dirname(sys.path[0])
+ self.session = get_session()
+ self.nova = nova_client(conf.nova_version, self.session)
+ self.compute_hosts = self.nova.hypervisors.list(detailed=True)
+ for host in self.compute_hosts:
+ host_dict = host.__dict__
+ self.compute_host = host_dict['hypervisor_hostname']
+ self.compute_ip = host_dict['host_ip']
+ tmp_sock = socket.socket(socket.AF_INET, socket.SOCK_DGRAM)
+ tmp_sock.connect(("8.8.8.8", 80))
+
+ ## control_ip is the IP of primary interface of control node i.e.
+ ## eth0, eno1. It is used by collectd monitor to communicate with
+ ## sample inspector.
+ ## TODO (umar) see if mgmt IP of control is a better option. Also
+ ## primary interface may not be the right option
+ self.control_ip = tmp_sock.getsockname()[0]
+ self.compute_user = getpass.getuser()
+ self.interface_name = os.environ.get('INTERFACE_NAME') or ''
+ self.inspector_type = os.environ.get('INSPECTOR_TYPE', 'sample')
+ self.auth_url = os.environ.get('OS_AUTH_URL')
+ self.username = os.environ.get('OS_USERNAME')
+ self.password = os.environ.get('OS_PASSWORD')
+ self.project_name = os.environ.get('OS_PROJECT_NAME')
+ self.user_domain_name = os.environ.get('OS_USER_DOMAIN_NAME') or 'default'
+ self.user_domain_id = os.environ.get('OS_USER_DOMAIN_ID')
+ self.project_domain_name = os.environ.get('OS_PROJECT_DOMAIN_NAME') or 'default'
+ self.project_domain_id = os.environ.get('OS_PROJECT_DOMAIN_ID')
+ self.ssh_opts_cpu = '-o UserKnownHostsFile=/dev/null -o StrictHostKeyChecking=no'
+
+ def start(self):
+ self.log.info("Collectd monitor start.........")
+ f = open("%s/tests/collectd.conf" % self.top_dir, 'w')
+ collectd_conf_file = """
+Hostname %s
+FQDNLookup false
+Interval 1
+MaxReadInterval 2
+
+<LoadPlugin python>
+Globals true
+</LoadPlugin>
+LoadPlugin ovs_events
+LoadPlugin logfile
+
+<Plugin logfile>
+ File \"/var/log/collectd.log\"
+ Timestamp true
+ LogLevel \"info\"
+</Plugin>
+
+<Plugin python>
+ ModulePath \"/home/%s\"
+ LogTraces true
+ Interactive false
+ Import \"collectd_plugin\"
+ <Module \"collectd_plugin\">
+ control_ip \"%s\"
+ compute_ip \"%s\"
+ compute_host \"%s\"
+ compute_user \"%s\"
+ inspector_type \"%s\"
+ os_auth_url \"%s\"
+ os_username \"%s\"
+ os_password \"%s\"
+ os_project_name \"%s\"
+ os_user_domain_name \"%s\"
+ os_user_domain_id \"%s\"
+ os_project_domain_name \"%s\"
+ os_project_domain_id \"%s\"
+ </Module>
+</Plugin>
+
+<Plugin ovs_events>
+ Port 6640
+ Socket \"/var/run/openvswitch/db.sock\"
+ Interfaces \"@INTERFACE_NAME@\"
+ SendNotification true
+ DispatchValues false
+</Plugin>
+ """ % (self.compute_host, self.compute_user, self.control_ip, self.compute_ip, self.compute_host, self.compute_user,
+ self.inspector_type, self.auth_url, self.username, self.password, self.project_name, self.user_domain_name,
+ self.user_domain_id, self.project_domain_name, self.project_domain_id)
+ f.write(collectd_conf_file)
+ f.close()
+
+ os.system(" scp %s %s/tests/collectd.conf %s@%s: " % (self.ssh_opts_cpu, self.top_dir, self.compute_user, self.compute_ip))
+ self.log.info("after first scp")
+ ## @TODO (umar) Always assuming that the interface is assigned an IP if
+ ## interface name is not provided. See if there is a better approach
+ os.system(""" ssh %s %s@%s \"if [ -n \"%s\" ]; then
+ dev=%s
+ else
+ dev=\$(sudo ip a | awk '/ %s\//{print \$NF}')
+ fi
+ sed -i -e \"s/@INTERFACE_NAME@/\$dev/\" collectd.conf
+ collectd_conf=/opt/collectd/etc/collectd.conf
+ if [ -e \$collectd_conf ]; then
+ sudo cp \$collectd_conf \${collectd_conf}-doctor-saved
+ else
+ sudo touch \${collectd_conf}-doctor-created
+ fi
+ sudo mv collectd.conf /opt/collectd/etc/collectd.conf\" """ % (self.ssh_opts_cpu, self.compute_user, self.compute_ip, self.interface_name, self.interface_name, self.compute_ip))
+ self.log.info("after first ssh")
+ os.system(" scp %s %s/tests/lib/monitors/collectd/collectd_plugin.py %s@%s:collectd_plugin.py " % (self.ssh_opts_cpu, self.top_dir, self.compute_user, self.compute_ip))
+ self.log.info("after sec scp")
+ os.system(" ssh %s %s@%s \"sudo pkill collectd; sudo /opt/collectd/sbin/collectd\" " % (self.ssh_opts_cpu, self.compute_user, self.compute_ip))
+ self.log.info("after sec ssh")
+
+ def stop(self):
+ os.system(" ssh %s %s@%s \"sudo pkill collectd\" " % (self.ssh_opts_cpu, self.compute_user, self.compute_ip))
+
+ def cleanup(self):
+ os.system(""" ssh %s %s@%s \"
+ collectd_conf=/opt/collectd/etc/collectd.conf
+ if [ -e \"\${collectd_conf}-doctor-created\" ]; then
+ sudo rm \"\${collectd_conf}-doctor-created\"
+ sudo rm \$collectd_conf
+ elif [ -e \"\${collectd_conf}-doctor-saved\" ]; then
+ sudo cp -f \"\${collectd_conf}-doctor-saved\" \$collectd_conf
+ sudo rm \"\${collectd_conf}-doctor-saved\"
+ fi\" """ % (self.ssh_opts_cpu, self.compute_user, self.compute_ip))
+ os.remove("%s/tests/collectd.conf" % self.top_dir)
done
}
-start_monitor() {
- pgrep -f "python monitor.py" && return 0
- sudo -E python monitor.py "$COMPUTE_HOST" "$COMPUTE_IP" "$INSPECTOR_TYPE" \
- > monitor.log 2>&1 &
-}
-
-stop_monitor() {
- pgrep -f "python monitor.py" || return 0
- sudo kill $(pgrep -f "python monitor.py")
-}
-
start_consumer() {
pgrep -f "python consumer.py" && return 0
python consumer.py "$CONSUMER_PORT" > consumer.log 2>&1 &
echo "disabling network of compute host [$COMPUTE_HOST] for 3 mins..."
cat > disable_network.sh << 'END_TXT'
#!/bin/bash -x
-dev=$(sudo ip a | awk '/ @COMPUTE_IP@\//{print $NF}')
sleep 1
+if [ -n "@INTERFACE_NAME@" ]; then
+ dev=@INTERFACE_NAME@
+else
+ dev=$(sudo ip a | awk '/ @COMPUTE_IP@\//{print $NF}')
+fi
sudo ip link set $dev down
echo "doctor set link down at" $(date "+%s.%N")
sleep 180
sleep 1
END_TXT
sed -i -e "s/@COMPUTE_IP@/$COMPUTE_IP/" disable_network.sh
+ sed -i -e "s/@INTERFACE_NAME@/$INTERFACE_NAME/" disable_network.sh
chmod +x disable_network.sh
scp $ssh_opts_cpu disable_network.sh "$COMPUTE_USER@$COMPUTE_IP:"
ssh $ssh_opts_cpu "$COMPUTE_USER@$COMPUTE_IP" 'nohup ./disable_network.sh > disable_network.log 2>&1 &'
wait_consumer 60
#keep 'at' as the last keyword just before the value, and
#use regex to get value instead of the fixed column
+ if [ ! -f monitor.log ]; then
+ scp $ssh_opts_cpu "$COMPUTE_USER@$COMPUTE_IP:monitor.log" .
+ fi
detected=$(grep "doctor monitor detected at" monitor.log |\
- sed -e "s/^.* at //")
+ sed -e "s/^.* at //" | tail -1)
notified=$(grep "doctor consumer notified at" consumer.log |\
sed -e "s/^.* at //" | tail -1)
cleanup() {
set +e
echo "cleanup..."
- stop_monitor
stop_inspector
stop_consumer
unset_forced_down_hosts
+ stop_monitor
collect_logs
vms=$(openstack $as_doctor_user server list)
cleanup_installer
cleanup_inspector
+ cleanup_monitor
# NOTE: Temporal log printer.
for f in $(find . -name '*.log')
source $TOP_DIR/functions-common
source $TOP_DIR/lib/installer
source $TOP_DIR/lib/inspector
+source $TOP_DIR/lib/monitor
+
+rm -f *.log
setup_installer
inject_failure
check_host_status "(DOWN|UNKNOWN)" 60
-calculate_notification_time
unset_forced_down_hosts
+calculate_notification_time
collect_logs
run_profiler
Code Review / doctor.git / commitdiff