import os
from os.path import isfile, join
import random
+import signal
import sys
import time
# creating test user...
self.user.create()
+
+ def setup_fault_management(self):
+ # user settings...
self.user.update_quota()
# creating VM...
self.alarm.create()
# starting doctor sample components...
- self.inspector.start()
+ # tbd tojuvone: move inspector and consumer to common setup
+ # when they support updating VMs via instance.create and
+ # instance.delete alarm
+ self.inspector.start()
+ self.consumer.start()
self.down_host = self.get_host_info_for_random_vm()
self.monitor.start(self.down_host)
- self.consumer.start()
-
- def run(self):
- """run doctor test"""
+ def test_fault_management(self):
try:
- LOG.info('doctor test starting.......')
+ LOG.info('doctor fault management test starting.......')
# prepare test env
- self.setup()
+ self.setup_fault_management()
# wait for aodh alarms are updated in caches for event evaluator,
# sleep time should be larger than event_alarm_cache_ttl
# (default 60)
- time.sleep(60)
+ # (tojuvone) Fraser currently needs 120
+ time.sleep(120)
# injecting host failure...
# NOTE (umar) add INTERFACE_NAME logic to host injection
notification_time = calculate_notification_time(LogFile)
if notification_time < 1 and notification_time > 0:
- LOG.info('doctor test successfully, notification_time=%s'
- % notification_time)
+ LOG.info('doctor fault management test successfully, '
+ 'notification_time=%s' % notification_time)
else:
- LOG.error('doctor test failed, notification_time=%s'
- % notification_time)
+ LOG.error('doctor fault management test failed, '
+ 'notification_time=%s' % notification_time)
sys.exit(1)
if self.conf.profiler_type:
- LOG.info('doctor test begin to run profile.......')
+ LOG.info('doctor fault management test begin to run '
+ 'profile.......')
self.collect_logs()
self.run_profiler()
+ except Exception as e:
+ LOG.error('doctor fault management test failed, '
+ 'Exception=%s' % e)
+ sys.exit(1)
+ finally:
+ self.cleanup_fault_management()
+
+ def _amount_compute_nodes(self):
+ services = self.nova.services.list(binary='nova-compute')
+ return len(services)
+
+ def test_maintenance(self):
+ cnodes = self._amount_compute_nodes()
+ if cnodes < 3:
+ # need 2 compute for redundancy and one spare to migrate
+ LOG.info('not enough compute nodes, skipping doctor '
+ 'maintenance test')
+ return
+ try:
+ LOG.info('doctor maintenance test starting.......')
+ # TODO (tojuvone) test setup and actual test
+ except Exception as e:
+ LOG.error('doctor maintenance test failed, Exception=%s' % e)
+ sys.exit(1)
+ # TODO (tojuvone) finally: test case specific cleanup
+
+ def run(self):
+ """run doctor tests"""
+ try:
+ LOG.info('doctor test starting.......')
+ # prepare common test env
+ self.setup()
+ if self.conf.test_case == 'all':
+ self.test_fault_management()
+ self.test_maintenance()
+ else:
+ function = 'test_%s' % self.conf.test_case
+ if hasattr(self, function):
+ getattr(self, function)()
+ else:
+ raise Exception('Can not find function <%s> in'
+ 'DoctorTest, see config manual'
+ % function)
except Exception as e:
LOG.error('doctor test failed, Exception=%s' % e)
sys.exit(1)
num = random.randint(0, self.conf.instance_count - 1)
vm_name = "%s%d" % (self.conf.instance_basename, num)
- servers = \
- {getattr(server, 'name'): server
- for server in self.nova.servers.list()}
+ servers = {getattr(server, 'name'): server
+ for server in self.nova.servers.list()}
server = servers.get(vm_name)
if not server:
- raise \
- Exception('Can not find instance: vm_name(%s)' % vm_name)
+ raise Exception('Can not find instance: vm_name(%s)' % vm_name)
host_name = server.__dict__.get('OS-EXT-SRV-ATTR:hypervisor_hostname')
host_ip = self.installer.get_host_ip_from_hostname(host_name)
# TODO(yujunz) check the actual delay to verify time sync status
# expected ~1s delay from $trigger to $linkdown
relative_start = linkdown
- os.environ['DOCTOR_PROFILER_T00'] = \
- str(int((linkdown - relative_start) * 1000))
- os.environ['DOCTOR_PROFILER_T01'] = \
- str(int((detected - relative_start) * 1000))
- os.environ['DOCTOR_PROFILER_T03'] = \
- str(int((vmdown - relative_start) * 1000))
- os.environ['DOCTOR_PROFILER_T04'] = \
- str(int((hostdown - relative_start) * 1000))
- os.environ['DOCTOR_PROFILER_T09'] = \
- str(int((notified - relative_start) * 1000))
+ os.environ['DOCTOR_PROFILER_T00'] = (
+ str(int((linkdown - relative_start) * 1000)))
+ os.environ['DOCTOR_PROFILER_T01'] = (
+ str(int((detected - relative_start) * 1000)))
+ os.environ['DOCTOR_PROFILER_T03'] = (
+ str(int((vmdown - relative_start) * 1000)))
+ os.environ['DOCTOR_PROFILER_T04'] = (
+ str(int((hostdown - relative_start) * 1000)))
+ os.environ['DOCTOR_PROFILER_T09'] = (
+ str(int((notified - relative_start) * 1000)))
profiler_main(log=LOG)
- def cleanup(self):
+ def cleanup_fault_management(self):
self.unset_forced_down_hosts()
self.inspector.stop()
self.monitor.stop()
self.consumer.stop()
- self.installer.cleanup()
self.alarm.delete()
self.instance.delete()
self.network.delete()
- self.image.delete()
self.fault.cleanup()
+
+ def cleanup(self):
+ self.installer.cleanup()
+ self.image.delete()
self.user.delete()
+ # Kill possible hanging subprocess
+ os.killpg(0, signal.SIGKILL)
def main():
"""doctor main"""
+ # TODO (tojuvone): JIRA DOCTOR-123: Test cases have some issue to always
+ # kill all subprocesses. To ensure they are killed this group is done so
+ # all processes can be killed without knowing what they are.
+ os.setpgrp()
test_dir = os.path.split(os.path.realpath(__file__))[0]
doctor_root_dir = os.path.dirname(test_dir)
Code Review / doctor.git / blobdiff