def setUp(self):
django.setup()
Scheduler.objects.initialize_scheduler()
self._driver = MagicMock()
scheduler_mgr.sync_with_database()
offer_mgr.clear()
self.node_agent_1 = 'agent_1'
self.node_agent_2 = 'agent_2'
self.slave_infos = [SlaveInfo('host_1', slave_id=self.node_agent_1),
SlaveInfo('host_2', slave_id=self.node_agent_2)]
node_mgr.clear()
node_mgr.register_agent_ids([self.node_agent_1, self.node_agent_2])
with patch('scheduler.node.manager.api.get_slaves') as mock_get_slaves:
mock_get_slaves.return_value = self.slave_infos
node_mgr.sync_with_database('master_host', 5050)
# Ignore initial cleanup tasks
for node in node_mgr.get_nodes():
node.initial_cleanup_completed()
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)
job_type_mgr.sync_with_database()
self._scheduling_thread = SchedulingThread(self._driver, '123')
def _perform_sync(self):
"""Performs the sync with the database
"""
scheduler_mgr.sync_with_database()
job_type_mgr.sync_with_database()
workspace_mgr.sync_with_database()
mesos_master = scheduler_mgr.mesos_address
node_mgr.sync_with_database(mesos_master.hostname, mesos_master.port)
self._sync_running_job_executions()
def test_paused_scheduler(self, mock_taskinfo):
"""Tests running the scheduling thread with a paused scheduler"""
mock_taskinfo.return_value = MagicMock()
offer_1 = ResourceOffer('offer_1', self.node_agent_1, NodeResources(cpus=2.0, mem=1024.0, disk=1024.0))
offer_2 = ResourceOffer('offer_2', self.node_agent_2, NodeResources(cpus=25.0, mem=2048.0, disk=2048.0))
offer_mgr.add_new_offers([offer_1, offer_2])
Scheduler.objects.update(is_paused=True)
scheduler_mgr.sync_with_database()
num_tasks = self._scheduling_thread._perform_scheduling()
self.assertEqual(num_tasks, 0)
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
job_type_mgr.sync_with_database()
scheduler_mgr.sync_with_database()
workspace_mgr.sync_with_database()
# Start up background threads
self._db_sync_thread = DatabaseSyncThread(self._driver)
db_sync_thread = threading.Thread(target=self._db_sync_thread.run)
db_sync_thread.daemon = True
db_sync_thread.start()
self._recon_thread = ReconciliationThread()
recon_thread = threading.Thread(target=self._recon_thread.run)
recon_thread.daemon = True
recon_thread.start()
self._scheduling_thread = SchedulingThread(self._driver, self._framework_id)
scheduling_thread = threading.Thread(target=self._scheduling_thread.run)
scheduling_thread.daemon = True
scheduling_thread.start()
self._task_thread = TaskUpdateThread()
task_thread = threading.Thread(target=self._task_thread.run)
task_thread.daemon = True
task_thread.start()
self._reconcile_running_jobs()
def test_paused_scheduler(self, mock_taskinfo):
"""Tests running the scheduling thread with a paused scheduler"""
mock_taskinfo.return_value = MagicMock()
offer_1 = ResourceOffer(
'offer_1', self.node_agent_1,
NodeResources(cpus=2.0, mem=1024.0, disk=1024.0))
offer_2 = ResourceOffer(
'offer_2', self.node_agent_2,
NodeResources(cpus=25.0, mem=2048.0, disk=2048.0))
offer_mgr.add_new_offers([offer_1, offer_2])
Scheduler.objects.update(is_paused=True)
scheduler_mgr.sync_with_database()
num_tasks = self._scheduling_thread._perform_scheduling()
self.assertEqual(num_tasks, 0)
def _perform_sync(self):
"""Performs the sync with the database
"""
scheduler_mgr.sync_with_database()
job_type_mgr.sync_with_database()
workspace_mgr.sync_with_database()
mesos_master = scheduler_mgr.mesos_address
node_mgr.sync_with_database(mesos_master.hostname, mesos_master.port)
# Kill running tasks for canceled job executions
for task_to_kill in job_exe_mgr.sync_with_database():
pb_task_to_kill = mesos_pb2.TaskID()
pb_task_to_kill.value = task_to_kill.id
logger.info('Killing task %s', task_to_kill.id)
self._driver.killTask(pb_task_to_kill)
if settings.SECRETS_URL:
secrets_mgr.sync_with_backend()
def setUp(self):
django.setup()
Scheduler.objects.initialize_scheduler()
self._driver = MagicMock()
scheduler_mgr.sync_with_database()
offer_mgr.clear()
self.node_agent_1 = 'agent_1'
self.node_agent_2 = 'agent_2'
self.slave_infos = [
SlaveInfo('host_1', slave_id=self.node_agent_1),
SlaveInfo('host_2', slave_id=self.node_agent_2)
]
node_mgr.clear()
node_mgr.register_agent_ids([self.node_agent_1, self.node_agent_2])
with patch('scheduler.node.manager.api.get_slaves') as mock_get_slaves:
mock_get_slaves.return_value = self.slave_infos
node_mgr.sync_with_database('master_host', 5050)
# Ignore initial cleanup tasks and health check tasks
for node in node_mgr.get_nodes():
node._last_heath_task = now()
node._initial_cleanup_completed()
node._update_state()
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)
job_type_mgr.sync_with_database()
self._scheduling_thread = SchedulingThread(self._driver, '123')
