doctor_tests/main.py

   1 ##############################################################################
   2 # Copyright (c) 2017 ZTE Corporation and others.
   3 #
   4 # All rights reserved. This program and the accompanying materials
   5 # are made available under the terms of the Apache License, Version 2.0
   6 # which accompanies this distribution, and is available at
   7 # http://www.apache.org/licenses/LICENSE-2.0
   8 ##############################################################################
   9 import os
  10 from os.path import isfile, join
  11 import random
  12 import signal
  13 import sys
  14 import time
  15
  16 from doctor_tests.alarm import Alarm
  17 from doctor_tests.common.constants import Host
  18 from doctor_tests.common.utils import match_rep_in_file
  19 from doctor_tests import config
  20 from doctor_tests.consumer import get_consumer
  21 from doctor_tests.identity_auth import get_identity_auth
  22 from doctor_tests.identity_auth import get_session
  23 from doctor_tests.image import Image
  24 from doctor_tests.instance import Instance
  25 from doctor_tests.inspector import get_inspector
  26 from doctor_tests.installer import get_installer
  27 import doctor_tests.logger as doctor_log
  28 from doctor_tests.network import Network
  29 from doctor_tests.monitor import get_monitor
  30 from doctor_tests.os_clients import nova_client
  31 from doctor_tests.profiler_poc import main as profiler_main
  32 from doctor_tests.scenario.common import calculate_notification_time
  33 from doctor_tests.scenario.network_failure import NetworkFault
  34 from doctor_tests.user import User
  35
  36
  37 Logger = doctor_log.Logger('doctor')
  38 LOG = Logger.getLogger()
  39 LogFile = Logger.getLogFilename()
  40
  41
  42 class DoctorTest(object):
  43
  44     def __init__(self, conf):
  45         self.conf = conf
  46         self.image = Image(self.conf, LOG)
  47         self.user = User(self.conf, LOG)
  48         self.network = Network(self.conf, LOG)
  49         self.instance = Instance(self.conf, LOG)
  50         self.alarm = Alarm(self.conf, LOG)
  51         self.installer = get_installer(self.conf, LOG)
  52         self.inspector = get_inspector(self.conf, LOG)
  53         self.monitor = get_monitor(self.conf,
  54                                    self.inspector.get_inspector_url(),
  55                                    LOG)
  56         self.consumer = get_consumer(self.conf, LOG)
  57         self.fault = NetworkFault(self.conf, self.installer, LOG)
  58         auth = get_identity_auth(project=self.conf.doctor_project)
  59         self.nova = nova_client(self.conf.nova_version,
  60                                 get_session(auth=auth))
  61         self.down_host = None
  62
  63     def setup(self):
  64         # prepare the cloud env
  65         self.installer.setup()
  66
  67         # preparing VM image...
  68         self.image.create()
  69
  70         # creating test user...
  71         self.user.create()
  72
  73     def setup_fault_management(self):
  74         # user settings...
  75         self.user.update_quota()
  76
  77         # creating VM...
  78         self.network.create()
  79         self.instance.create()
  80         self.instance.wait_for_vm_launch()
  81
  82         # creating alarm...
  83         self.alarm.create()
  84
  85         # starting doctor sample components...
  86         # tbd tojuvone: move inspector and consumer to common setup
  87         # when they support updating VMs via instance.create and
  88         # instance.delete alarm
  89
  90         self.inspector.start()
  91         self.consumer.start()
  92         self.down_host = self.get_host_info_for_random_vm()
  93         self.monitor.start(self.down_host)
  94
  95     def test_fault_management(self):
  96         try:
  97             LOG.info('doctor fault management test starting.......')
  98
  99             # prepare test env
 100             self.setup_fault_management()
 101
 102             # wait for aodh alarms are updated in caches for event evaluator,
 103             # sleep time should be larger than event_alarm_cache_ttl
 104             # (default 60)
 105             time.sleep(60)
 106
 107             # injecting host failure...
 108             # NOTE (umar) add INTERFACE_NAME logic to host injection
 109
 110             self.fault.start(self.down_host)
 111             time.sleep(10)
 112
 113             # verify the test results
 114             # NOTE (umar) copy remote monitor.log file when monitor=collectd
 115             self.check_host_status(self.down_host.name, 'down')
 116
 117             notification_time = calculate_notification_time(LogFile)
 118             if notification_time < 1 and notification_time > 0:
 119                 LOG.info('doctor fault management test successfully, '
 120                          'notification_time=%s' % notification_time)
 121             else:
 122                 LOG.error('doctor fault management test failed, '
 123                           'notification_time=%s' % notification_time)
 124                 sys.exit(1)
 125
 126             if self.conf.profiler_type:
 127                 LOG.info('doctor fault management test begin to run '
 128                          'profile.......')
 129                 self.collect_logs()
 130                 self.run_profiler()
 131         except Exception as e:
 132             LOG.error('doctor fault management test failed, '
 133                       'Exception=%s' % e)
 134             sys.exit(1)
 135         finally:
 136             self.cleanup_fault_management()
 137
 138     def _amount_compute_nodes(self):
 139         services = self.nova.services.list(binary='nova-compute')
 140         return len(services)
 141
 142     def test_maintenance(self):
 143         cnodes = self._amount_compute_nodes()
 144         if cnodes < 3:
 145             # need 2 compute for redundancy and one spare to migrate
 146             LOG.info('not enough compute nodes, skipping doctor '
 147                      'maintenance test')
 148             return
 149         try:
 150             LOG.info('doctor maintenance test starting.......')
 151             # TODO (tojuvone) test setup and actual test
 152         except Exception as e:
 153             LOG.error('doctor maintenance test failed, Exception=%s' % e)
 154             sys.exit(1)
 155         # TODO (tojuvone) finally: test case specific cleanup
 156
 157     def run(self):
 158         """run doctor tests"""
 159         try:
 160             LOG.info('doctor test starting.......')
 161             # prepare common test env
 162             self.setup()
 163             if self.conf.test_case == 'all':
 164                 self.test_fault_management()
 165                 self.test_maintenance()
 166             else:
 167                 function = 'test_%s' % self.conf.test_case
 168                 if hasattr(self, function):
 169                     getattr(self, function)()
 170                 else:
 171                     raise Exception('Can not find function <%s> in'
 172                                     'DoctorTest, see config manual'
 173                                     % function)
 174         except Exception as e:
 175             LOG.error('doctor test failed, Exception=%s' % e)
 176             sys.exit(1)
 177         finally:
 178             self.cleanup()
 179
 180     def get_host_info_for_random_vm(self):
 181         num = random.randint(0, self.conf.instance_count - 1)
 182         vm_name = "%s%d" % (self.conf.instance_basename, num)
 183
 184         servers = {getattr(server, 'name'): server
 185                    for server in self.nova.servers.list()}
 186         server = servers.get(vm_name)
 187         if not server:
 188             raise Exception('Can not find instance: vm_name(%s)' % vm_name)
 189         host_name = server.__dict__.get('OS-EXT-SRV-ATTR:hypervisor_hostname')
 190         host_ip = self.installer.get_host_ip_from_hostname(host_name)
 191
 192         LOG.info('Get host info(name:%s, ip:%s) which vm(%s) launched at'
 193                  % (host_name, host_ip, vm_name))
 194         return Host(host_name, host_ip)
 195
 196     def check_host_status(self, hostname, state):
 197         service = self.nova.services.list(host=hostname,
 198                                           binary='nova-compute')
 199         host_state = service[0].__dict__.get('state')
 200         assert host_state == state
 201
 202     def unset_forced_down_hosts(self):
 203         if self.down_host:
 204             self.nova.services.force_down(self.down_host.name,
 205                                           'nova-compute', False)
 206             time.sleep(2)
 207             self.check_host_status(self.down_host.name, 'up')
 208
 209     def collect_logs(self):
 210         self.fault.get_disable_network_log()
 211
 212     def run_profiler(self):
 213
 214         net_down_log_file = self.fault.get_disable_network_log()
 215         reg = '(?<=doctor set link down at )\d+.\d+'
 216         linkdown = float(match_rep_in_file(reg, net_down_log_file).group(0))
 217
 218         reg = '(.* doctor mark vm.* error at )(\d+.\d+)'
 219         vmdown = float(match_rep_in_file(reg, LogFile).group(2))
 220
 221         reg = '(.* doctor mark host.* down at )(\d+.\d+)'
 222         hostdown = float(match_rep_in_file(reg, LogFile).group(2))
 223
 224         reg = '(?<=doctor monitor detected at )\d+.\d+'
 225         detected = float(match_rep_in_file(reg, LogFile).group(0))
 226
 227         reg = '(?<=doctor consumer notified at )\d+.\d+'
 228         notified = float(match_rep_in_file(reg, LogFile).group(0))
 229
 230         # TODO(yujunz) check the actual delay to verify time sync status
 231         # expected ~1s delay from $trigger to $linkdown
 232         relative_start = linkdown
 233         os.environ['DOCTOR_PROFILER_T00'] = (
 234             str(int((linkdown - relative_start) * 1000)))
 235         os.environ['DOCTOR_PROFILER_T01'] = (
 236             str(int((detected - relative_start) * 1000)))
 237         os.environ['DOCTOR_PROFILER_T03'] = (
 238             str(int((vmdown - relative_start) * 1000)))
 239         os.environ['DOCTOR_PROFILER_T04'] = (
 240             str(int((hostdown - relative_start) * 1000)))
 241         os.environ['DOCTOR_PROFILER_T09'] = (
 242             str(int((notified - relative_start) * 1000)))
 243
 244         profiler_main(log=LOG)
 245
 246     def cleanup_fault_management(self):
 247         self.unset_forced_down_hosts()
 248         self.inspector.stop()
 249         self.monitor.stop()
 250         self.consumer.stop()
 251         self.alarm.delete()
 252         self.instance.delete()
 253         self.network.delete()
 254         self.fault.cleanup()
 255
 256     def cleanup(self):
 257         self.installer.cleanup()
 258         self.image.delete()
 259         self.user.delete()
 260         # Kill possible hanging subprocess
 261         os.killpg(0, signal.SIGKILL)
 262
 263
 264 def main():
 265     """doctor main"""
 266     # TODO (tojuvone): JIRA DOCTOR-123: Test cases have some issue to always
 267     # kill all subprocesses. To ensure they are killed this group is done so
 268     # all processes can be killed without knowing what they are.
 269     os.setpgrp()
 270     test_dir = os.path.split(os.path.realpath(__file__))[0]
 271     doctor_root_dir = os.path.dirname(test_dir)
 272
 273     config_file_dir = '{0}/{1}'.format(doctor_root_dir, 'etc/')
 274     config_files = [join(config_file_dir, f)
 275                     for f in os.listdir(config_file_dir)
 276                     if isfile(join(config_file_dir, f))]
 277
 278     conf = config.prepare_conf(args=sys.argv[1:],
 279                                config_files=config_files)
 280
 281     doctor = DoctorTest(conf)
 282     doctor.run()