def setUp(self):
django.setup()
self.job_exe = job_test_utils.create_job_exe()
# Clear error cache so tests work correctly
reset_error_cache()
def setUp(self):
django.setup()
self.error_1 = error_test_utils.create_error(name='error_1',
category='SYSTEM')
self.error_2 = error_test_utils.create_error(name='error_2',
category='SYSTEM')
self.error_3 = error_test_utils.create_error(name='error_3',
category='ALGORITHM')
# Clear error cache so tests work correctly
reset_error_cache()
def setUp(self):
django.setup()
reset_error_cache()
self.framework_id = '1234'
Scheduler.objects.initialize_scheduler()
Scheduler.objects.update(
num_message_handlers=0
) # Prevent message handler tasks from scheduling
self._client = MagicMock()
scheduler_mgr.sync_with_database()
scheduler_mgr.update_from_mesos(framework_id=self.framework_id)
resource_mgr.clear()
job_exe_mgr.clear()
self.agent_1 = Agent('agent_1', 'host_1')
self.agent_2 = Agent('agent_2', 'host_2')
self.agent_3 = Agent('agent_3', 'host_2')
node_mgr.clear()
node_mgr.register_agents([self.agent_1, self.agent_2])
node_mgr.sync_with_database(scheduler_mgr.config)
# Ignore initial cleanup, health check, and image pull tasks
for node in node_mgr.get_nodes():
node._last_health_task = now()
node._initial_cleanup_completed()
node._is_image_pulled = True
node._update_state()
if node.agent_id == 'agent_1':
self.node_1_id = node.id
cleanup_mgr.update_nodes(node_mgr.get_nodes())
self.node_1 = Node.objects.get(id=self.node_1_id)
# Ignore system tasks
system_task_mgr._is_db_update_completed = True
self.queue_1 = queue_test_utils.create_queue(cpus_required=4.0,
mem_required=1024.0,
disk_in_required=100.0,
disk_out_required=200.0,
disk_total_required=300.0)
self.queue_2 = queue_test_utils.create_queue(cpus_required=8.0,
mem_required=512.0,
disk_in_required=400.0,
disk_out_required=45.0,
disk_total_required=445.0)
self.queue_large = queue_test_utils.create_queue(
resources=NodeResources([Cpus(
125.0), Mem(12048.0), Disk(12048.0)]))
job_type_mgr.sync_with_database()
def setUp(self):
django.setup()
# Clear error cache so tests work correctly
reset_error_cache()
self.agent_id = 'agent'
self.node_model_1 = node_test_utils.create_node()
self.job_exe_1 = job_test_utils.create_running_job_exe(agent_id=self.agent_id, node=self.node_model_1)
self.node_model_2 = node_test_utils.create_node()
self.job_exe_2 = job_test_utils.create_running_job_exe(agent_id=self.agent_id, node=self.node_model_2)
self.task_mgr = TaskManager()
self.job_exe_mgr = JobExecutionManager()
def test_get_error_missing_default(self):
"""Tests that custom error is returned when a non-registered name is found in the mapping"""
error_interface_dict = {
'version': '1.0',
'exit_codes': {
'1': self.error_1.name,
'2': self.error_2.name,
'3': self.error_3.name,
},
}
default_error = error_test_utils.create_error()
default_error.is_builtin = True
default_error.save()
# Reset error cache so tests work correctly
reset_error_cache()
error_interface = ErrorInterface(error_interface_dict)
error = error_interface.get_error(4, default_error.name)
self.assertIsNotNone(error)
self.assertEqual(error.name, default_error.name)
def setUp(self):
django.setup()
# Clear error cache so tests work correctly
reset_error_cache()
def setUp(self):
django.setup()
reset_error_cache()
def setUp(self):
django.setup()
reset_error_cache()
self.task_mgr = TaskManager()
def initialize(self):
"""Initializes the scheduler and gets it ready to connect to Mesos. This method should only ever be called once.
"""
initialize_system()
# Initial database sync
logger.info('Performing initial sync with Scale database')
logger.info('Retrieving errors...')
reset_error_cache()
logger.info('Retrieving job execution metrics...')
job_exe_mgr.init_with_database()
logger.info('Retrieving job types...')
job_type_mgr.sync_with_database()
logger.info('Retrieving workspaces...')
workspace_mgr.sync_with_database()
logger.info('Retrieving scheduler settings...')
scheduler_mgr.sync_with_database()
# Start up background threads
self._threads = []
logger.info('Starting up background threads')
self._messaging_thread = MessagingThread()
restart_msg = RestartScheduler()
restart_msg.when = now()
self._messaging_thread.add_initial_messages([restart_msg])
messaging_thread = threading.Thread(target=self._messaging_thread.run)
messaging_thread.daemon = True
messaging_thread.start()
self._threads.append(messaging_thread)
self._recon_thread = ReconciliationThread()
recon_thread = threading.Thread(target=self._recon_thread.run)
recon_thread.daemon = True
recon_thread.start()
self._threads.append(recon_thread)
self._scheduler_status_thread = SchedulerStatusThread()
scheduler_status_thread = threading.Thread(target=self._scheduler_status_thread.run)
scheduler_status_thread.daemon = True
scheduler_status_thread.start()
self._threads.append(scheduler_status_thread)
self._scheduling_thread = SchedulingThread(self._client)
scheduling_thread = threading.Thread(target=self._scheduling_thread.run)
scheduling_thread.daemon = True
scheduling_thread.start()
self._threads.append(scheduling_thread)
self._sync_thread = SyncThread(self._driver)
sync_thread = threading.Thread(target=self._sync_thread.run)
sync_thread.daemon = True
sync_thread.start()
self._threads.append(sync_thread)
self._task_handling_thread = TaskHandlingThread(self._driver)
task_handling_thread = threading.Thread(target=self._task_handling_thread.run)
task_handling_thread.daemon = True
task_handling_thread.start()
self._threads.append(task_handling_thread)
self._task_update_thread = TaskUpdateThread()
task_update_thread = threading.Thread(target=self._task_update_thread.run)
task_update_thread.daemon = True
task_update_thread.start()
self._threads.append(task_update_thread)
def registered(self, driver, frameworkId, masterInfo):
"""
Invoked when the scheduler successfully registers with a Mesos master.
It is called with the frameworkId, a unique ID generated by the
master, and the masterInfo which is information about the master
itself.
See documentation for :meth:`mesos_api.mesos.Scheduler.registered`.
"""
self._driver = driver
self._framework_id = frameworkId.value
self._master_hostname = masterInfo.hostname
self._master_port = masterInfo.port
logger.info(
'Scale scheduler registered as framework %s with Mesos master at %s:%i',
self._framework_id, self._master_hostname, self._master_port)
initialize_system()
Scheduler.objects.update_master(self._master_hostname,
self._master_port)
scheduler_mgr.update_from_mesos(
self._framework_id,
HostAddress(self._master_hostname, self._master_port))
recon_mgr.driver = self._driver
# Initial database sync
logger.info('Performing initial sync with Scale database')
reset_error_cache()
job_exe_mgr.init_with_database()
job_type_mgr.sync_with_database()
scheduler_mgr.sync_with_database()
workspace_mgr.sync_with_database()
# Start up background threads
self._messaging_thread = MessagingThread()
restart_msg = RestartScheduler()
restart_msg.when = now()
self._messaging_thread.add_initial_messages([restart_msg])
messaging_thread = threading.Thread(target=self._messaging_thread.run)
messaging_thread.daemon = True
messaging_thread.start()
self._recon_thread = ReconciliationThread()
recon_thread = threading.Thread(target=self._recon_thread.run)
recon_thread.daemon = True
recon_thread.start()
self._scheduler_status_thread = SchedulerStatusThread()
scheduler_status_thread = threading.Thread(
target=self._scheduler_status_thread.run)
scheduler_status_thread.daemon = True
scheduler_status_thread.start()
self._scheduling_thread = SchedulingThread(self._driver)
scheduling_thread = threading.Thread(
target=self._scheduling_thread.run)
scheduling_thread.daemon = True
scheduling_thread.start()
self._sync_thread = SyncThread(self._driver)
sync_thread = threading.Thread(target=self._sync_thread.run)
sync_thread.daemon = True
sync_thread.start()
self._task_handling_thread = TaskHandlingThread(self._driver)
task_handling_thread = threading.Thread(
target=self._task_handling_thread.run)
task_handling_thread.daemon = True
task_handling_thread.start()
self._task_update_thread = TaskUpdateThread()
task_update_thread = threading.Thread(
target=self._task_update_thread.run)
task_update_thread.daemon = True
task_update_thread.start()
self._reconcile_running_jobs()
