def guard(self):
logger.info(_('{} guarding cluster at {}').format(current_process().name, self.pid))
self.start_event.set()
Stat(self).save()
logger.info(_('Q Cluster-{} running.').format(self.parent_pid))
scheduler(broker=self.broker)
counter = 0
cycle = Conf.GUARD_CYCLE # guard loop sleep in seconds
# Guard loop. Runs at least once
while not self.stop_event.is_set() or not counter:
# Check Workers
for p in self.pool:
# Are you alive?
if not p.is_alive() or (self.timeout and p.timer.value == 0):
self.reincarnate(p)
continue
# Decrement timer if work is being done
if self.timeout and p.timer.value > 0:
p.timer.value -= cycle
# Check Monitor
if not self.monitor.is_alive():
self.reincarnate(self.monitor)
# Check Pusher
if not self.pusher.is_alive():
self.reincarnate(self.pusher)
# Call scheduler once a minute (or so)
counter += cycle
if counter >= 30 and Conf.SCHEDULER:
counter = 0
scheduler(broker=self.broker)
# Save current status
Stat(self).save()
sleep(cycle)
self.stop()
def guard(self):
logger.info(
_(f"{current_process().name} guarding cluster {humanize(self.cluster_id.hex)}"
))
self.start_event.set()
Stat(self).save()
logger.info(_(f"Q Cluster {humanize(self.cluster_id.hex)} running."))
counter = 0
cycle = Conf.GUARD_CYCLE # guard loop sleep in seconds
# Guard loop. Runs at least once
while not self.stop_event.is_set() or not counter:
# Check Workers
for p in self.pool:
with p.timer.get_lock():
# Check if worker is alive or timed out
if not p.is_alive() or p.timer.value == 0:
self.reincarnate_worker(p)
continue
# Decrement timer if work is being done
if p.timer.value > 0:
p.timer.value = 0 if p.timer.value < cycle else p.timer.value - cycle
# Check Monitor
if not self.monitor.is_alive():
self.reincarnate_monitor()
# Check Pusher
if not self.pusher.is_alive():
self.reincarnate_pusher()
counter += cycle
# Run every 30 seconds
if counter >= 30:
# Update cluster heartbeat time
now_time = timezone.now()
self.cluster_model.heartbeat_time = now_time
self.cluster_model.save(update_fields=['heartbeat_time'])
# Clean up other cluster instances that have died
self.cluster_model.filter(heartbeat_time__lt=now_time -
timedelta(minutes=1)).delete()
# Call scheduler
if Conf.SCHEDULER:
scheduler(broker=self.broker)
counter = 0
# Save current status
Stat(self).save()
sleep(cycle)
self.stop()
def test_bad_secret(broker, monkeypatch):
broker.list_key = "test_bad_secret:q"
async_task("math.copysign", 1, -1, broker=broker)
stop_event = Event()
stop_event.set()
start_event = Event()
cluster_id = uuidlib.uuid4()
s = Sentinel(stop_event,
start_event,
cluster_id=cluster_id,
broker=broker,
start=False)
Stat(s).save()
# change the SECRET
monkeypatch.setattr(Conf, "SECRET_KEY", "OOPS")
stat = Stat.get_all()
assert len(stat) == 0
assert Stat.get(pid=s.parent_pid, cluster_id=cluster_id) is None
task_queue = Queue()
pusher(task_queue, stop_event, broker=broker)
result_queue = Queue()
task_queue.put("STOP")
worker(
task_queue,
result_queue,
Value("f", -1),
)
assert result_queue.qsize() == 0
broker.delete_queue()
def test_bad_broker(broker, mocker):
mocker.patch.object(broker, 'set_stat',
side_effect=Exception('Unusable connection'))
stop_event = Event()
stop_event.set()
start_event = Event()
s = Sentinel(stop_event, start_event, broker=broker, start=False)
mock_close = mocker.patch.object(broker, 'close')
Stat(s).save()
assert mock_close.called
def stop(self):
Stat(self).save()
name = current_process().name
logger.info(_('{} stopping cluster processes').format(name))
# Stopping pusher
self.event_out.set()
# Wait for it to stop
while self.pusher.is_alive():
sleep(0.1)
Stat(self).save()
# Put poison pills in the queue
for __ in range(len(self.pool)):
self.task_queue.put('STOP')
self.task_queue.close()
# wait for the task queue to empty
self.task_queue.join_thread()
# Wait for all the workers to exit
while len(self.pool):
for p in self.pool:
if not p.is_alive():
self.pool.remove(p)
sleep(0.1)
Stat(self).save()
# Finally stop the monitor
self.result_queue.put('STOP')
self.result_queue.close()
# Wait for the result queue to empty
self.result_queue.join_thread()
logger.info(_('{} waiting for the monitor.').format(name))
# Wait for everything to close or time out
count = 0
if not self.timeout:
self.timeout = 30
while self.status() == Conf.STOPPING and count < self.timeout * 10:
sleep(0.1)
Stat(self).save()
count += 1
# Final status
Stat(self).save()
def spawn_cluster(self):
self.pool = []
Stat(self).save()
db.connection.close()
# spawn worker pool
for __ in range(self.pool_size):
self.spawn_worker()
# spawn auxiliary
self.monitor = self.spawn_monitor()
self.pusher = self.spawn_pusher()
# set worker cpu affinity if needed
if psutil and Conf.CPU_AFFINITY:
set_cpu_affinity(Conf.CPU_AFFINITY, [w.pid for w in self.pool])
def guard(self):
logger.info(
_(
f"{current_process().name} guarding cluster {humanize(self.cluster_id.hex)}"
)
)
self.start_event.set()
Stat(self).save()
logger.info(_(f"Q Cluster {humanize(self.cluster_id.hex)} running."))
counter = 0
cycle = Conf.GUARD_CYCLE # guard loop sleep in seconds
# Guard loop. Runs at least once
while not self.stop_event.is_set() or not counter:
# Check Workers
for p in self.pool:
with p.timer.get_lock():
# Are you alive?
if not p.is_alive() or p.timer.value == 0:
self.reincarnate(p)
continue
# Decrement timer if work is being done
if p.timer.value > 0:
p.timer.value -= cycle
# Check Monitor
if not self.monitor.is_alive():
self.reincarnate(self.monitor)
# Check Pusher
if not self.pusher.is_alive():
self.reincarnate(self.pusher)
# Call scheduler once a minute (or so)
counter += cycle
if counter >= 30 and Conf.SCHEDULER:
counter = 0
scheduler(broker=self.broker)
# Save current status
Stat(self).save()
sleep(cycle)
self.stop()
def test_bad_secret(broker, monkeypatch):
broker.list_key = 'test_bad_secret:q'
async('math.copysign', 1, -1, broker=broker)
stop_event = Event()
stop_event.set()
start_event = Event()
s = Sentinel(stop_event, start_event, broker=broker, start=False)
Stat(s).save()
# change the SECRET
monkeypatch.setattr(Conf, "SECRET_KEY", "OOPS")
stat = Stat.get_all()
assert len(stat) == 0
assert Stat.get(s.parent_pid) is None
task_queue = Queue()
pusher(task_queue, stop_event, broker=broker)
result_queue = Queue()
task_queue.put('STOP')
worker(task_queue, result_queue, Value('f', -1), )
assert result_queue.qsize() == 0
broker.delete_queue()
