def test_job_execution_monitoring(self):
watcher = SchedulerWatcher(self.scheduler)
self.scheduler.add_job(
lambda: time.sleep(0.02),
id='waiting_job',
name='Waiting job',
jobstore='in_memory',
trigger='interval',
seconds=0.2,
next_run_time=datetime.now()
)
job_events = watcher.jobs['waiting_job']['events']
self.assertEqual(1, len(job_events))
self.assertEqual('job_added', job_events[0]['event_name'])
time.sleep(0.05)
self.assertEqual(3, len(job_events), 'Job execution needs to be tracked in job events')
self.assertEqual(
'job_submitted',
job_events[1]['event_name'],
'Job submision needs to be tracked in job events'
)
self.assertEqual('job_executed', job_events[2]['event_name'], 'Job execution needs to be tracked in job events')
time.sleep(0.2)
self.assertEqual(5, len(job_events), 'Subsequent executions get tracked')
def test_job_properties_on_add(self):
watcher = SchedulerWatcher(self.scheduler)
self.scheduler.add_job(
lambda x, y: x + y,
id='added_job',
name='Added job',
jobstore='in_memory',
trigger='interval',
minutes=60,
args=(1,),
kwargs={'y': 2}
)
self.assertIn('added_job', watcher.jobs)
job_properties = watcher.jobs['added_job']['properties']
self.assertEqual('added_job', job_properties['id'], 'Job properties should have the job id')
self.assertEqual('Added job', job_properties['name'], 'Job properties should have the job name')
self.assertIn('trigger', job_properties, 'Job properties should have a representation of the trigger')
self.assertEqual('in_memory', job_properties['jobstore'], 'Job properties should have the jobstore name')
self.assertEqual('default', job_properties['executor'], 'Job properties should have the executor name')
self.assertIn('lambda', job_properties['func'], 'Job properties should have the function string repr')
self.assertIn('func_ref', job_properties, 'Job properties should have the function reference')
self.assertEqual('(1,)', job_properties['args'], 'Job properties should have the job arguments')
self.assertEqual("{'y': 2}", job_properties['kwargs'], 'Job properties should have the job keyword arguments')
self.assertIn('pending', job_properties, 'Job properties should have the job pending status')
self.assertFalse(job_properties['pending'], 'Job status should not be pending')
self.assertIn('coalesce', job_properties, 'Job properties should have the job coalesce configuration')
self.assertIn('next_run_time', job_properties, 'Job properties should have the next run time calculated')
self.assertIn('misfire_grace_time', job_properties, 'Job properties should have the misfire grace time')
self.assertIn('max_instances', job_properties, 'Job properties should have the max instances configuration')
def __init__(self, scheduler, capabilities=None, operation_timeout=1):
self.scheduler = scheduler
patch_scheduler(scheduler)
self.capabilities = {
'pause_job': False,
'remove_job': False,
'pause_scheduler': False,
'stop_scheduler': False,
'run_job': False,
}
if not (isinstance(operation_timeout, int)
or isinstance(operation_timeout, float)):
raise TypeError(
'operation_timeout should be either an int or a float')
if operation_timeout <= 0:
raise ValueError('operation_timeout should be a positive number')
self.operation_timeout = operation_timeout
if capabilities is not None:
if isinstance(capabilities, dict):
self.capabilities.update(capabilities)
else:
raise TypeError(
'capabilities should be a dict of str -> bool pairs')
self._scheduler_listener = SchedulerWatcher(scheduler)
self._web_server = flask.Flask(__name__)
self._socket_io = None
try:
# TODO: see if we can support eventlet in the future.
self._socket_io = flask_socketio.SocketIO(self._web_server,
async_mode='gevent')
except ValueError:
self._socket_io = flask_socketio.SocketIO(self._web_server,
async_mode='threading')
self._init_endpoints()
self._web_server_thread = None
self._scheduler_lock = threading.Lock()
def test_watcher_injection(self):
watcher = SchedulerWatcher(self.scheduler)
self.assertEqual(watcher.scheduler, self.scheduler, 'Watcher should keep a reference to the scheduler')
self.assertEqual(1, len(self.scheduler._listeners), 'Watcher should inject itself as a scheduler listener')
self.assertEqual(
self.scheduler._listeners[0][1], EVENT_ALL, 'Watcher should register iself to watch all events'
)
def test_removed_jobs_are_only_flagged_as_removed(self):
self.scheduler.add_job(lambda: 0, id='a_job')
watcher = SchedulerWatcher(self.scheduler)
self.assertIn('a_job', watcher.jobs)
self.assertIsNone(watcher.jobs['a_job']['removed_time'])
self.scheduler.remove_job('a_job')
self.assertIn('a_job', watcher.jobs, 'removed jobs should be still tracked in the scheduler watcher')
self.assertIsNotNone(watcher.jobs['a_job']['removed_time'], 'removed_time should be set')
def test_job_event_history_is_limited(self):
watcher = SchedulerWatcher(self.scheduler, max_events_per_job=4)
self.scheduler.add_job(lambda: 0, trigger='interval', seconds=0.01, id='recurrent_job')
time.sleep(0.1)
# recurrent_job should have been executed ~10 times now, generating ~20 events (submission + execution).
self.assertEqual(
watcher.max_events_per_job,
len(watcher.jobs['recurrent_job']['events']),
'job event history should be limited'
)
def test_adding_and_removing_executors(self, mock_notify_executor_event):
watcher = SchedulerWatcher(self.scheduler)
self.scheduler.add_executor(ThreadPoolExecutor(), alias='new_executor')
self.assertIn('new_executor', watcher.executors)
mock_notify_executor_event.assert_called()
mock_notify_executor_event.reset_mock()
self.scheduler.remove_executor('new_executor')
self.assertNotIn('new_executor', watcher.executors)
mock_notify_executor_event.assert_called()
def test_scheduler_summary(self):
watcher = SchedulerWatcher(self.scheduler)
summary = watcher.scheduler_summary()
self.assertEqual(sorted(['scheduler', 'jobs', 'executors', 'jobstores']), sorted(summary.keys()))
self.assertEqual('running', summary['scheduler']['state'], 'scheduler_summary should have the scheduler status')
self.assertEqual(2, len(summary['executors']), 'scheduler_summaru should have the two added executors')
self.assertEqual(2, len(summary['jobstores']), 'scheduler_summary should have the two executors')
self.assertEqual(0, len(summary['jobs']), 'scheduler_summary should have no jobs')
self.scheduler.add_job(lambda: 0, id='job_1')
summary = watcher.scheduler_summary()
self.assertIn('job_1', summary['jobs'], 'scheduler_summary should have the added jobs in it')
self.scheduler.remove_job('job_1')
summary = watcher.scheduler_summary()
self.assertIn('job_1', summary['jobs'], 'scheduler_summary should have all jobs in it, even if job was removed')
def test_removing_all_jobs_flags_all_as_removed(self, mock_notify_job_event):
watcher = SchedulerWatcher(self.scheduler)
self.scheduler.add_job(lambda: 0, id='job_1', jobstore='default', trigger='interval', minutes=60)
self.scheduler.add_job(lambda: 0, id='job_2', jobstore='in_memory', trigger='interval', minutes=60)
self.assertEqual(2, len(watcher.jobs))
self.assertEqual(2, mock_notify_job_event.call_count)
mock_notify_job_event.reset_mock()
self.scheduler.remove_all_jobs()
self.assertEqual(2, len(watcher.jobs), 'job count should not change after removing all jobs')
self.assertEqual(2, mock_notify_job_event.call_count)
def test_adding_a_jobstore_adds_all_jobs_in_it(self, mock_notify_jobstore_event, mock_notify_job_event, _):
watcher = SchedulerWatcher(self.scheduler)
jobstore = MemoryJobStore()
jobstore.add_job(Job(scheduler=self.scheduler, id='job_1', next_run_time=datetime.now() + timedelta(days=1)))
jobstore.add_job(Job(scheduler=self.scheduler, id='job_2', next_run_time=datetime.now() + timedelta(days=2)))
self.assertEqual(0, len(watcher.jobs))
self.scheduler.add_jobstore(jobstore, alias='in_memory_2')
self.assertIn('in_memory_2', watcher.jobstores, 'Watcher should have the new jobstore tracked')
self.assertEqual(2, len(watcher.jobs), 'Watcher should add all jobs in the newly added jobstore')
self.assertTrue(all([job_id in watcher.jobs for job_id in ['job_1', 'job_2']]))
self.assertEqual(2, mock_notify_job_event.call_count)
mock_notify_jobstore_event.assert_called_once()
def test_removing_a_jobstore_removes_all_jobs(self, mock_notify_jobstore_event):
watcher = SchedulerWatcher(self.scheduler)
self.scheduler.add_job(lambda: 0, id='job_1', jobstore='in_memory', trigger='interval', minutes=60)
self.scheduler.add_job(lambda: 0, id='job_2', jobstore='in_memory', trigger='interval', minutes=60)
self.assertEqual(2, len(watcher.jobs))
self.assertIsNone(watcher.jobs['job_1']['removed_time'], 'job_1 removed time should be None')
self.assertEqual('in_memory', watcher.jobs['job_1']['properties']['jobstore'])
self.scheduler.remove_jobstore('in_memory')
mock_notify_jobstore_event.assert_called()
self.assertEqual(2, len(watcher.jobs), 'The amount of jobs after removing a jobstore should not change')
self.assertIsNotNone(watcher.jobs['job_1']['removed_time'], 'job_1 removed time should be set')
self.assertIsNotNone(watcher.jobs['job_2']['removed_time'], 'job_2 removed time should be set')
def test_job_inspection_matches_job_added_event(self):
# We're going to add two jobs that should have the exact same properties, except for the id, in two different
# stages of the usage: before the watcher is created and after we start watching for events.
def job_function(x, y):
return x + y
next_run_time = datetime.now() + timedelta(hours=1)
# Job that is added before the user calls us.
self.scheduler.add_job(
job_function,
id='job_1',
name='Added job',
jobstore='in_memory',
trigger='interval',
minutes=60,
args=(1,),
kwargs={'y': 2},
next_run_time=next_run_time
)
watcher = SchedulerWatcher(self.scheduler)
# Job that gets added after we start watching.
self.scheduler.add_job(
job_function,
id='job_2',
name='Added job',
jobstore='in_memory',
trigger='interval',
minutes=60,
args=(1,),
kwargs={'y': 2},
next_run_time=next_run_time
)
self.assertEqual(2, len(watcher.jobs))
job_1 = watcher.jobs['job_1']
job_2 = watcher.jobs['job_2']
for property_name in job_1['properties'].keys():
# All properties, except the id, should match.
if property_name == 'id':
continue
self.assertEqual(job_1['properties'][property_name], job_2['properties'][property_name])
def test_job_failure_monitoring(self):
watcher = SchedulerWatcher(self.scheduler)
def fail():
time.sleep(0.02)
return 0 / 0
self.scheduler.add_job(
fail,
id='failing_job',
name='Failing job',
jobstore='in_memory',
trigger='interval',
next_run_time=datetime.now(),
minutes=60
)
failing_job_events = watcher.jobs['failing_job']['events']
time.sleep(0.05)
self.assertEqual(3, len(failing_job_events))
self.assertEqual('job_error', failing_job_events[2]['event_name'])
def test_scheduler_inspection(self):
self.scheduler.add_job(lambda: 0, jobstore='in_memory', trigger='interval', minutes=60, id='test_job')
watcher = SchedulerWatcher(self.scheduler)
self.assertEqual('running', watcher.scheduler_info['state'], 'Watcher should inspect scheduler status')
self.assertEqual(
str(self.scheduler.timezone),
watcher.scheduler_info['timezone'],
'Watcher should inspect scheduler timezone'
)
self.assertEqual(
'BackgroundScheduler', watcher.scheduler_info['class'], 'Watcher should inspect scheduler class'
)
self.assertEqual(2, len(watcher.jobstores), 'Watcher should inspect all scheduler jobstores')
self.assertIn('in_memory', watcher.jobstores, 'Watcher should have inspected the in_memory jobstore')
self.assertEqual(2, len(watcher.executors), 'Watcher should inspect all scheduler executors')
self.assertIn('secondary_executor', watcher.executors, 'Watcher should have inspected the secondary_executor')
self.assertEqual(1, len(watcher.jobs), 'Watcher should inspect all jobs in scheduler on init')
self.assertIn('test_job', watcher.jobs, 'Watcher should index jobs by id')
def test_modified_job_properties_are_tracked(self):
self.scheduler.add_job(
lambda x, y: x + y,
id='a_job',
name='A job',
jobstore='in_memory',
trigger='interval',
minutes=60,
args=(1,),
kwargs={'y': 2}
)
watcher = SchedulerWatcher(self.scheduler)
self.assertEqual(watcher.jobs['a_job']['modified_time'], watcher.jobs['a_job']['added_time'])
next_run_time = watcher.jobs['a_job']['properties']['next_run_time'][0]
self.scheduler.modify_job('a_job', name='A modified job', next_run_time=datetime.now() + timedelta(days=1))
self.assertGreater(watcher.jobs['a_job']['modified_time'], watcher.jobs['a_job']['added_time'])
self.assertEqual('A modified job', watcher.jobs['a_job']['properties']['name'])
self.assertGreater(watcher.jobs['a_job']['properties']['next_run_time'][0], next_run_time)
class SchedulerUI(SchedulerEventsListener):
"""
A web server that monitors your scheduler and serves a web application to visualize events.
By the default the server web application served is a view-only UI, but by enabling capabilities it may be allowed
to control the scheduler and its jobs.
Args:
scheduler (apscheduler.schedulers.base.BaseScheduler):
The scheduler to monitor.
capabilities (dict):
(Optional)
A dictionary of the capabilities to enable in the server and client. By default the UI is view-only.
Supported capabilities:
* Pause/Resume Scheduler: set `pause_scheduler` to :data:`True`.
* Pause/Resume Jobs: set `pause_job` to :data:`True`.
* Remove Jobs: set `remove_job` to :data:True.
operation_timeout (float):
(Optional) The amount of seconds to wait for the serializing lock when performing actions on the
scheduler or on its jobs from the UI.
Basic Usage:
>>> from apscheduler.schedulers.background import BackgroundScheduler
>>> from apschedulerui.web import SchedulerUI
>>> scheduler = BackgroundScheduler()
>>> ui = SchedulerUI(scheduler)
>>> ui.start() # Server available at localhost:5000.
Configuring capabilities:
>>> ui = SchedulerUI(scheduler, capabilities={'pause_scheduler': True}) # All omitted capabilities are False.
>>> ui = SchedulerUI(scheduler, capabilities={'pause_job': True, 'remove_job': True})
"""
def __init__(self, scheduler, capabilities=None, operation_timeout=1):
self.scheduler = scheduler
patch_scheduler(scheduler)
self.capabilities = {
'pause_job': False,
'remove_job': False,
'pause_scheduler': False,
'stop_scheduler': False,
'run_job': False,
}
if not (isinstance(operation_timeout, int)
or isinstance(operation_timeout, float)):
raise TypeError(
'operation_timeout should be either an int or a float')
if operation_timeout <= 0:
raise ValueError('operation_timeout should be a positive number')
self.operation_timeout = operation_timeout
if capabilities is not None:
if isinstance(capabilities, dict):
self.capabilities.update(capabilities)
else:
raise TypeError(
'capabilities should be a dict of str -> bool pairs')
self._scheduler_listener = SchedulerWatcher(scheduler)
self._web_server = flask.Flask(__name__)
self._socket_io = None
try:
# TODO: see if we can support eventlet in the future.
self._socket_io = flask_socketio.SocketIO(self._web_server,
async_mode='gevent')
except ValueError:
self._socket_io = flask_socketio.SocketIO(self._web_server,
async_mode='threading')
self._init_endpoints()
self._web_server_thread = None
self._scheduler_lock = threading.Lock()
def start(self, host='0.0.0.0', port=5000, daemon=True):
"""
Starts listening for events from the scheduler and starts the web server that serves the UI in a new thread.
Args:
host (str):
(Optional) The address to bind the web server to. Default `0.0.0.0`.
port (int):
(Optional) The port to which the web server will bind. Defaults to :data:`5000`.
daemon (bool):
(Optional) If :data:`True` (default) starts the server as daemon.
"""
self._scheduler_listener.add_listener(self)
self._web_server_thread = threading.Thread(target=self._start,
name='apscheduler-ui',
args=(host, port))
self._web_server_thread.daemon = daemon
self._web_server_thread.start()
def _init_endpoints(self):
if self.capabilities.get('pause_scheduler', False):
self._web_server.add_url_rule('/api/scheduler/pause',
'pause_scheduler',
self._pause_scheduler,
methods=['POST'])
self._web_server.add_url_rule('/api/scheduler/resume',
'resume_scheduler',
self._resume_scheduler,
methods=['POST'])
if self.capabilities.get('stop_scheduler', False):
self._web_server.add_url_rule('/api/scheduler/stop',
'stop_scheduler',
self._stop_scheduler,
methods=['POST'])
self._web_server.add_url_rule('/api/scheduler/start',
'start_scheduler',
self._start_scheduler,
methods=['POST'])
if self.capabilities.get('remove_job', False):
self._web_server.add_url_rule('/api/job/<job_id>/remove',
'remove_job',
self._remove_job,
methods=['POST'])
if self.capabilities.get('pause_job', False):
self._web_server.add_url_rule('/api/job/<job_id>/pause',
'pause_job',
self._pause_job,
methods=['POST'])
self._web_server.add_url_rule('/api/job/<job_id>/resume',
'resume_job',
self._resume_job,
methods=['POST'])
if self.capabilities.get('run_job', False):
self._web_server.add_url_rule('/api/job/<job_id>/run_now',
'run_job',
self._run_job,
methods=['POST'])
self._web_server.add_url_rule('/',
'index',
self._index,
defaults={'path': ''})
self._web_server.add_url_rule('/<path:path>', 'index', self._index)
self._socket_io.on_event('connected', self._client_connected)
def _index(self, path):
return self._web_server.send_static_file('index.html')
def _exec_scheduler_command(self, func, *args, **kwargs):
if self._scheduler_lock.acquire(timeout=self.operation_timeout):
try:
func(*args, **kwargs)
return 'ok'
except JobLookupError:
flask.abort(404, description="Job not found")
finally:
self._scheduler_lock.release()
else:
flask.abort(
408,
description=
"Failed to acquire scheduler lock to perform operation")
def _pause_scheduler(self):
return self._exec_scheduler_command(self.scheduler.pause)
def _resume_scheduler(self):
return self._exec_scheduler_command(self.scheduler.resume)
def _stop_scheduler(self):
return self._exec_scheduler_command(self.scheduler.shutdown,
wait=False)
def _start_scheduler(self):
return self._exec_scheduler_command(self.scheduler.start)
def _pause_job(self, job_id):
return self._exec_scheduler_command(self.scheduler.pause_job, job_id)
def _resume_job(self, job_id):
return self._exec_scheduler_command(self.scheduler.resume_job, job_id)
def _run_job(self, job_id, next_run_time=None):
logging.getLogger('apschedulerui').info('Running job %s' % job_id)
if not job_id:
return Response(status=404)
if not next_run_time:
next_run_time = datetime.now()
def _run_job_impl():
job = self.scheduler.get_job(job_id)
if not job:
raise JobLookupError(job_id)
# If a job is periodic (has an interval trigger) it should be triggered by modifying the trigger it already
# has. Otherwise, it can be rescheduled to be ran now.
if isinstance(job.trigger, IntervalTrigger) or isinstance(
job.trigger, CronTrigger):
self.scheduler.modify_job(job_id, next_run_time=next_run_time)
else:
job.reschedule(trigger='date', run_date=next_run_time)
return self._exec_scheduler_command(_run_job_impl)
def _remove_job(self, job_id):
return self._exec_scheduler_command(self.scheduler.remove_job, job_id)
def _client_connected(self):
logging.getLogger('apschedulerui').debug('Client connected')
flask_socketio.emit('init_jobs',
self._scheduler_listener.scheduler_summary())
flask_socketio.emit('init_capabilities', self.capabilities)
def _job_event(self, event):
self._socket_io.emit('job_event', event)
def _scheduler_event(self, event):
self._socket_io.emit('scheduler_event', event)
def _jobstore_event(self, event):
self._socket_io.emit('jobstore_event', event)
def _executor_event(self, event):
self._socket_io.emit('executor_event', event)
def _start(self, host, port):
self._socket_io.run(self._web_server, host=host, port=port)
