class Consumer(object):
def __init__(self, invoker, config):
self.invoker = invoker
self.config = config
self.logfile = config.LOGFILE
self.loglevel = config.LOGLEVEL
self.threads = config.THREADS
self.periodic_commands = config.PERIODIC
self.default_delay = config.INITIAL_DELAY
self.backoff_factor = config.BACKOFF
self.max_delay = config.MAX_DELAY
self.utc = config.UTC
# initialize delay
self.delay = self.default_delay
self.logger = self.get_logger()
# queue to track messages to be processed
self._queue = IterableQueue()
self._pool = threading.BoundedSemaphore(self.threads)
self._shutdown = threading.Event()
# initialize the command schedule
self.schedule = CommandSchedule(self.invoker)
def get_logger(self):
log = logging.getLogger('huey.consumer.logger')
log.setLevel(self.loglevel)
if not log.handlers and self.logfile:
handler = RotatingFileHandler(self.logfile, maxBytes=1024*1024, backupCount=3)
handler.setFormatter(logging.Formatter("%(asctime)s:%(name)s:%(levelname)s:%(message)s"))
log.addHandler(handler)
return log
def get_now(self):
if self.utc:
return datetime.datetime.utcnow()
else:
return datetime.datetime.now()
def spawn(self, func, *args, **kwargs):
t = threading.Thread(target=func, args=args, kwargs=kwargs)
t.daemon = True
t.start()
return t
def start_periodic_task_scheduler(self):
self.logger.info('starting periodic task scheduler thread')
return self.spawn(self.schedule_periodic_tasks)
def schedule_periodic_tasks(self):
while not self._shutdown.is_set():
start = time.time()
self.enqueue_periodic_commands(self.get_now())
time.sleep(60 - (time.time() - start))
def start_scheduler(self):
self.logger.info('starting scheduler thread')
return self.spawn(self.schedule_commands)
def schedule_commands(self):
while not self._shutdown.is_set():
start = time.time()
self.check_schedule(self.get_now())
# check schedule once per second
delta = time.time() - start
if delta < 1:
time.sleep(1 - (time.time() - start))
def check_schedule(self, dt):
for command in self.schedule.commands():
if self.schedule.should_run(command, dt):
self.schedule.remove(command)
self.invoker.enqueue(command)
def enqueue_periodic_commands(self, dt):
self.logger.debug('enqueueing periodic commands')
for command in registry.get_periodic_commands():
if command.validate_datetime(dt):
self.invoker.enqueue(command)
def start_message_receiver(self):
self.logger.info('starting message receiver thread')
return self.spawn(self.receive_messages)
def receive_messages(self):
while not self._shutdown.is_set():
self.check_message()
def check_message(self):
try:
command = self.invoker.dequeue()
except QueueReadException:
self.logger.error('error reading from queue', exc_info=1)
except QueueException:
self.logger.error('queue exception', exc_info=1)
else:
if not command and not self.invoker.blocking:
# no new messages and our queue doesn't block, so sleep a bit before
# checking again
self.sleep()
elif command:
if self.schedule.should_run(command, self.get_now()):
self.process_command(command)
else:
# schedule the command for execution
self.schedule.add(command)
def process_command(self, command):
self._pool.acquire()
self.logger.info('processing: %s' % command)
self.delay = self.default_delay
# put the command into the queue for the worker pool
self._queue.put(command)
# wait to acknowledge receipt of the command
self.logger.debug('waiting for receipt of command')
self._queue.join()
def sleep(self):
if self.delay > self.max_delay:
self.delay = self.max_delay
self.logger.debug('no commands, sleeping for: %s' % self.delay)
time.sleep(self.delay)
self.delay *= self.backoff_factor
def start_worker_pool(self):
self.logger.info('starting worker threadpool')
return self.spawn(self.worker_pool)
def worker_pool(self):
for job in self._queue:
# spin up a worker with the given job
self.spawn(self.worker, job)
# indicate receipt of the task
self._queue.task_done()
def worker(self, command):
try:
self.invoker.execute(command)
except DataStorePutException:
self.logger.warn('error storing result', exc_info=1)
except:
self.logger.error('unhandled exception in worker thread', exc_info=1)
if command.retries:
self.requeue_command(command)
finally:
self._pool.release()
def requeue_command(self, command):
command.retries -= 1
self.logger.info('re-enqueueing task %s, %s tries left' % (command.task_id, command.retries))
try:
if command.retry_delay:
delay = datetime.timedelta(seconds=command.retry_delay)
command.execute_time = self.get_now() + delay
self.schedule.add(command)
else:
self.invoker.enqueue(command)
except QueueWriteException:
self.logger.error('unable to re-enqueue %s, error writing to backend' % command.task_id)
def start(self):
self.load_schedule()
self.start_scheduler()
if self.periodic_commands:
self.start_periodic_task_scheduler()
self._receiver_t = self.start_message_receiver()
self._worker_pool_t = self.start_worker_pool()
def shutdown(self):
self._shutdown.set()
self._queue.put(StopIteration)
def handle_signal(self, sig_num, frame):
self.logger.info('received SIGTERM, shutting down')
self.shutdown()
def set_signal_handler(self):
self.logger.info('setting signal handler')
signal.signal(signal.SIGTERM, self.handle_signal)
def load_schedule(self):
self.logger.info('loading command schedule')
self.schedule.load()
def save_schedule(self):
self.logger.info('saving command schedule')
self.schedule.save()
def log_registered_commands(self):
msg = ['huey initialized with following commands']
for command in registry._registry:
msg.append('+ %s' % command)
self.logger.info('\n'.join(msg))
def run(self):
self.set_signal_handler()
self.log_registered_commands()
try:
self.start()
# it seems that calling self._shutdown.wait() here prevents the
# signal handler from executing
while not self._shutdown.is_set():
self._shutdown.wait(.1)
except:
self.logger.error('error', exc_info=1)
self.shutdown()
self.save_schedule()
self.logger.info('shutdown...')
def test_task_store(self):
schedule = CommandSchedule(res_invoker)
dt1 = datetime.datetime(2011, 1, 1, 0, 0)
dt2 = datetime.datetime(2035, 1, 1, 0, 0)
add2.schedule(args=('k', 'v'), eta=dt1)
cmd1 = res_invoker.dequeue()
add2.schedule(args=('k2', 'v2'), eta=dt2)
cmd2 = res_invoker.dequeue()
add2('k3', 'v3')
cmd3 = res_invoker.dequeue()
# add the command to the schedule
schedule.add(cmd1)
self.assertEqual(schedule.commands(), [cmd1])
# cmd1 is in the schedule, cmd2 is not
self.assertTrue(schedule.is_pending(cmd1))
self.assertFalse(schedule.is_pending(cmd2))
# multiple adds do not result in the list growing
schedule.add(cmd1)
self.assertEqual(schedule.commands(), [cmd1])
# removing a non-existant cmd does not error
schedule.remove(cmd2)
# add a future-dated command
schedule.add(cmd2)
# sanity check what should be run
self.assertTrue(schedule.should_run(cmd1))
self.assertFalse(schedule.should_run(cmd2))
self.assertTrue(schedule.should_run(cmd3))
# check remove works
schedule.remove(cmd1)
self.assertEqual(schedule.commands(), [cmd2])
# check saving works
schedule.save()
schedule._schedule = {}
self.assertEqual(schedule.commands(), [])
# check loading works
schedule.load()
self.assertEqual(schedule.commands(), [cmd2])
schedule.add(cmd1)
schedule.add(cmd3)
schedule.save()
schedule._schedule = {}
schedule.load()
self.assertEqual(len(schedule.commands()), 3)
self.assertTrue(cmd1 in schedule.commands())
self.assertTrue(cmd2 in schedule.commands())
self.assertTrue(cmd3 in schedule.commands())
def test_task_store(self):
schedule = CommandSchedule(res_invoker)
dt1 = datetime.datetime(2011, 1, 1, 0, 0)
dt2 = datetime.datetime(2035, 1, 1, 0, 0)
add2.schedule(args=('k', 'v'), eta=dt1)
cmd1 = res_invoker.dequeue()
add2.schedule(args=('k2', 'v2'), eta=dt2)
cmd2 = res_invoker.dequeue()
add2('k3', 'v3')
cmd3 = res_invoker.dequeue()
# add the command to the schedule
schedule.add(cmd1)
self.assertEqual(schedule.commands(), [cmd1])
# cmd1 is in the schedule, cmd2 is not
self.assertTrue(schedule.is_pending(cmd1))
self.assertFalse(schedule.is_pending(cmd2))
# multiple adds do not result in the list growing
schedule.add(cmd1)
self.assertEqual(schedule.commands(), [cmd1])
# removing a non-existant cmd does not error
schedule.remove(cmd2)
# add a future-dated command
schedule.add(cmd2)
# sanity check what should be run
self.assertTrue(schedule.should_run(cmd1))
self.assertFalse(schedule.should_run(cmd2))
self.assertTrue(schedule.should_run(cmd3))
# check remove works
schedule.remove(cmd1)
self.assertEqual(schedule.commands(), [cmd2])
# check saving works
schedule.save()
schedule._schedule = {}
self.assertEqual(schedule.commands(), [])
# check loading works
schedule.load()
self.assertEqual(schedule.commands(), [cmd2])
schedule.add(cmd1)
schedule.add(cmd3)
schedule.save()
schedule._schedule = {}
schedule.load()
self.assertEqual(len(schedule.commands()), 3)
self.assertTrue(cmd1 in schedule.commands())
self.assertTrue(cmd2 in schedule.commands())
self.assertTrue(cmd3 in schedule.commands())
class Consumer(object):
def __init__(self, invoker, config):
self.invoker = invoker
self.config = config
self.logfile = config.LOGFILE
self.loglevel = config.LOGLEVEL
self.threads = config.THREADS
self.periodic_commands = config.PERIODIC
self.default_delay = config.INITIAL_DELAY
self.backoff_factor = config.BACKOFF
self.max_delay = config.MAX_DELAY
self.utc = config.UTC
# initialize delay
self.delay = self.default_delay
self.logger = self.get_logger()
# queue to track messages to be processed
self._queue = IterableQueue()
self._pool = threading.BoundedSemaphore(self.threads)
self._shutdown = threading.Event()
# initialize the command schedule
self.schedule = CommandSchedule(self.invoker)
def get_logger(self):
log = logging.getLogger('huey.consumer.logger')
log.setLevel(self.loglevel)
if not log.handlers and self.logfile:
handler = RotatingFileHandler(self.logfile,
maxBytes=1024 * 1024,
backupCount=3)
handler.setFormatter(
logging.Formatter(
"%(asctime)s:%(name)s:%(levelname)s:%(message)s"))
log.addHandler(handler)
return log
def get_now(self):
if self.utc:
return datetime.datetime.utcnow()
else:
return datetime.datetime.now()
def spawn(self, func, *args, **kwargs):
t = threading.Thread(target=func, args=args, kwargs=kwargs)
t.daemon = True
t.start()
return t
def start_periodic_task_scheduler(self):
self.logger.info('starting periodic task scheduler thread')
return self.spawn(self.schedule_periodic_tasks)
def schedule_periodic_tasks(self):
while not self._shutdown.is_set():
start = time.time()
self.enqueue_periodic_commands(self.get_now())
time.sleep(60 - (time.time() - start))
def start_scheduler(self):
self.logger.info('starting scheduler thread')
return self.spawn(self.schedule_commands)
def schedule_commands(self):
while not self._shutdown.is_set():
start = time.time()
self.check_schedule(self.get_now())
# check schedule once per second
delta = time.time() - start
if delta < 1:
time.sleep(1 - (time.time() - start))
def check_schedule(self, dt):
for command in self.schedule.commands():
if self.schedule.should_run(command, dt):
self.schedule.remove(command)
self.invoker.enqueue(command)
def enqueue_periodic_commands(self, dt):
self.logger.debug('enqueueing periodic commands')
for command in registry.get_periodic_commands():
if command.validate_datetime(dt):
self.invoker.enqueue(command)
def start_message_receiver(self):
self.logger.info('starting message receiver thread')
return self.spawn(self.receive_messages)
def receive_messages(self):
while not self._shutdown.is_set():
self.check_message()
def check_message(self):
try:
command = self.invoker.dequeue()
except QueueReadException:
self.logger.error('error reading from queue', exc_info=1)
except QueueException:
self.logger.error('queue exception', exc_info=1)
else:
if not command and not self.invoker.blocking:
# no new messages and our queue doesn't block, so sleep a bit before
# checking again
self.sleep()
elif command:
if self.schedule.should_run(command, self.get_now()):
self.process_command(command)
else:
# schedule the command for execution
self.schedule.add(command)
def process_command(self, command):
self._pool.acquire()
self.logger.info('processing: %s' % command)
self.delay = self.default_delay
# put the command into the queue for the worker pool
self._queue.put(command)
# wait to acknowledge receipt of the command
self.logger.debug('waiting for receipt of command')
self._queue.join()
def sleep(self):
if self.delay > self.max_delay:
self.delay = self.max_delay
self.logger.debug('no commands, sleeping for: %s' % self.delay)
time.sleep(self.delay)
self.delay *= self.backoff_factor
def start_worker_pool(self):
self.logger.info('starting worker threadpool')
return self.spawn(self.worker_pool)
def worker_pool(self):
for job in self._queue:
# spin up a worker with the given job
self.spawn(self.worker, job)
# indicate receipt of the task
self._queue.task_done()
def worker(self, command):
try:
self.invoker.execute(command)
except DataStorePutException:
self.logger.warn('error storing result', exc_info=1)
except:
self.logger.error('unhandled exception in worker thread',
exc_info=1)
if command.retries:
self.requeue_command(command)
finally:
self._pool.release()
def requeue_command(self, command):
command.retries -= 1
self.logger.info('re-enqueueing task %s, %s tries left' %
(command.task_id, command.retries))
try:
if command.retry_delay:
delay = datetime.timedelta(seconds=command.retry_delay)
command.execute_time = self.get_now() + delay
self.schedule.add(command)
else:
self.invoker.enqueue(command)
except QueueWriteException:
self.logger.error(
'unable to re-enqueue %s, error writing to backend' %
command.task_id)
def start(self):
self.load_schedule()
self.start_scheduler()
if self.periodic_commands:
self.start_periodic_task_scheduler()
self._receiver_t = self.start_message_receiver()
self._worker_pool_t = self.start_worker_pool()
def shutdown(self):
self._shutdown.set()
self._queue.put(StopIteration)
def handle_signal(self, sig_num, frame):
self.logger.info('received SIGTERM, shutting down')
self.shutdown()
def set_signal_handler(self):
self.logger.info('setting signal handler')
signal.signal(signal.SIGTERM, self.handle_signal)
def load_schedule(self):
self.logger.info('loading command schedule')
self.schedule.load()
def save_schedule(self):
self.logger.info('saving command schedule')
self.schedule.save()
def log_registered_commands(self):
msg = ['huey initialized with following commands']
for command in registry._registry:
msg.append('+ %s' % command)
self.logger.info('\n'.join(msg))
def run(self):
self.set_signal_handler()
self.log_registered_commands()
try:
self.start()
# it seems that calling self._shutdown.wait() here prevents the
# signal handler from executing
while not self._shutdown.is_set():
self._shutdown.wait(.1)
except:
self.logger.error('error', exc_info=1)
self.shutdown()
self.save_schedule()
self.logger.info('shutdown...')