refactor the monitor

JIRA: DOCTOR-99

Change-Id: I1831d329ffe80435532678fcb23d54f310422ce8
Signed-off-by: dongwenjuan <dong.wenjuan@zte.com.cn>

diff --git a/tests/config.py b/tests/config.py
index 969d829..88910f6 100644 (file)
--- a/tests/config.py
+++ b/tests/config.py
@@ -16,14 +16,16 @@ import image
 import instance\r
 import network\r
 import inspector\r
+import monitor\r
 import os_clients\r
 import user\r
 \r
 \r
 def list_opts():\r
     return [\r
-        ('consumer', consumer.OPTS),\r
+        ('monitor', monitor.OPTS),\r
         ('inspector', inspector.OPTS),\r
+        ('consumer', consumer.OPTS),\r
         ('DEFAULT', itertools.chain(\r
             os_clients.OPTS,\r
             image.OPTS,\r

diff --git a/tests/inspector/__init__.py b/tests/inspector/__init__.py
index 35bdb5b..151cc03 100644 (file)
--- a/tests/inspector/__init__.py
+++ b/tests/inspector/__init__.py
@@ -11,6 +11,7 @@ import os
 from oslo_config import cfg
 from oslo_utils import importutils
 
+
 OPTS = [
     cfg.StrOpt('type',
                default=os.environ.get('INSPECTOR_TYPE', 'sample'),
@@ -19,7 +20,7 @@ OPTS = [
                required=True),
     cfg.StrOpt('ip',
                default='127.0.0.1',
-               help='the ip of default inspector',
+               help='the host ip of inspector',
                required=False),
     cfg.StrOpt('port',
                default='12345',
@@ -32,6 +33,7 @@ _inspector_name_class_mapping = {
     'sample': 'inspector.sample.SampleInspector',
 }
 
+
 def get_inspector(conf, log):
     inspector_class = _inspector_name_class_mapping[conf.inspector.type]
     return importutils.import_object(inspector_class, conf, log)

diff --git a/tests/main.py b/tests/main.py
index 797e28b..6644b54 100644 (file)
--- a/tests/main.py
+++ b/tests/main.py
@@ -18,6 +18,8 @@ from inspector import get_inspector
 import logger as doctor_log
 from user import User
 from network import Network
+from monitor import get_monitor
+
 
 LOG = doctor_log.Logger('doctor').getLogger()
 
@@ -32,6 +34,9 @@ class DoctorTest(object):
         self.instance = Instance(self.conf, LOG)
         self.alarm = Alarm(self.conf, LOG)
         self.inspector = get_inspector(self.conf, LOG)
+        self.monitor = get_monitor(self.conf,
+                                   self.inspector.get_inspector_url(),
+                                   LOG)
 
     def setup(self):
         # prepare the cloud env
@@ -53,6 +58,7 @@ class DoctorTest(object):
 
         # starting doctor sample components...
         self.inspector.start()
+        self.monitor.start()
 
     def run(self):
         """run doctor test"""
@@ -78,6 +84,7 @@ class DoctorTest(object):
         self.image.delete()
         self.user.delete()
         self.inspector.stop()
+        self.monitor.stop()
 
 
 def main():

diff --git a/tests/monitor/__init__.py b/tests/monitor/__init__.py
new file mode 100644 (file)
index 0000000..51a6a65
--- /dev/null
+++ b/tests/monitor/__init__.py
@@ -0,0 +1,28 @@
+##############################################################################
+# Copyright (c) 2017 ZTE 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
+##############################################################################
+from oslo_config import cfg
+from oslo_utils import importutils
+
+OPTS = [
+    cfg.StrOpt('type',
+               default='sample',
+               choices=['sample'],
+               help='the type of doctor monitor component',
+               required=True),
+]
+
+
+_monitor_name_class_mapping = {
+    'sample': 'monitor.sample.SampleMonitor'
+}
+
+def get_monitor(conf, inspector_url, log):
+    monitor_class = _monitor_name_class_mapping.get(conf.monitor.type)
+    return importutils.import_object(monitor_class, conf,
+                                     inspector_url, log)

diff --git a/tests/monitor/base.py b/tests/monitor/base.py
new file mode 100644 (file)
index 0000000..ccb647c
--- /dev/null
+++ b/tests/monitor/base.py
@@ -0,0 +1,27 @@
+##############################################################################
+# Copyright (c) 2017 ZTE 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 abc
+import six
+
+
+@six.add_metaclass(abc.ABCMeta)
+class BaseMonitor(object):
+    """Monitor computer fault and report error to the inspector"""
+    def __init__(self, conf, inspector_url, log):
+        self.conf = conf
+        self.log = log
+        self.inspector_url = inspector_url
+
+    @abc.abstractmethod
+    def start(self):
+        pass
+
+    @abc.abstractmethod
+    def stop(self):
+        pass

diff --git a/tests/monitor/sample.py b/tests/monitor/sample.py
new file mode 100644 (file)
index 0000000..4f1ce49
--- /dev/null
+++ b/tests/monitor/sample.py
@@ -0,0 +1,119 @@
+##############################################################################
+# Copyright (c) 2017 ZTE 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
+##############################################################################
+from datetime import datetime
+import json
+import requests
+import socket
+from threading import Thread
+import time
+
+from identity_auth import get_session
+from os_clients import nova_client
+from monitor.base import BaseMonitor
+
+
+class SampleMonitor(BaseMonitor):
+    event_type = "compute.host.down"
+
+    def __init__(self, conf, inspector_url, log):
+        super(SampleMonitor, self).__init__(conf, inspector_url, log)
+        self.session = get_session()
+        self.nova = nova_client(conf.nova_version, self.session)
+        self.hosts = self.nova.hypervisors.list(detailed=True)
+        self.pingers = []
+
+    def start(self):
+        self.log.info('sample monitor start......')
+        for host in self.hosts:
+            host_dict = host.__dict__
+            host_name = host_dict['hypervisor_hostname']
+            host_ip = host_dict['host_ip']
+            pinger = Pinger(host_name, host_ip, self, self.log)
+            pinger.start()
+            self.pingers.append(pinger)
+
+    def stop(self):
+        self.log.info('sample monitor stop......')
+        for pinger in self.pingers:
+            pinger.stop()
+            pinger.join()
+        del self.pingers
+
+    def report_error(self, hostname):
+        self.log.info('sample monitor report error......')
+        data = [
+            {
+                'id': 'monitor_sample_id1',
+                'time': datetime.now().isoformat(),
+                'type': self.event_type,
+                'details': {
+                    'hostname': hostname,
+                    'status': 'down',
+                    'monitor': 'monitor_sample',
+                    'monitor_event_id': 'monitor_sample_event1'
+                },
+            },
+        ]
+
+        auth_token = self.session.get_token() if \
+                     self.conf.inspector.type != 'sample' else None
+        headers = {
+            'Content-Type': 'application/json',
+            'Accept': 'application/json',
+            'X-Auth-Token': auth_token,
+        }
+
+        url = '%s%s' % (self.inspector_url, 'events') \
+            if self.inspector_url.endswith('/') else \
+            '%s%s' % (self.inspector_url, '/events')
+        requests.put(url, data=json.dumps(data), headers=headers)
+
+
+class Pinger(Thread):
+    interval = 0.1  # second
+    timeout = 0.1   # second
+    ICMP_ECHO_MESSAGE = '\x08\x00\xf7\xff\x00\x00\x00\x00'
+
+    def __init__(self, host_name, host_ip, monitor, log):
+        Thread.__init__(self)
+        self.monitor = monitor
+        self.hostname = host_name
+        self.ip_addr = host_ip or socket.gethostbyname(self.hostname)
+        self.log = log
+        self._stopped = False
+
+    def run(self):
+        while True:
+            if self._stopped:
+                return
+            self._run()
+            time.sleep(self.interval)
+
+    def stop(self):
+        self.log.info("Stopping Pinger host_name(%s), host_ip(%s)"
+                      % (self.hostname, self.ip_addr))
+        self._stopped = True
+
+    def _run(self):
+        self.log.info("Starting Pinger host_name(%s), host_ip(%s)"
+                      % (self.hostname, self.ip_addr))
+
+        sock = socket.socket(socket.AF_INET, socket.SOCK_RAW,
+                             socket.IPPROTO_ICMP)
+        sock.settimeout(self.timeout)
+        while True:
+            try:
+                sock.sendto(self.ICMP_ECHO_MESSAGE.encode(), (self.ip_addr, 0))
+                sock.recv(4096)
+            except socket.timeout:
+                self.log.info("doctor monitor detected at %s" % time.time())
+                self.monitor.report_error(self.hostname)
+                self.log.info("ping timeout, quit monitoring...")
+                self._stopped = True
+                return