Develop Reference

Join timeout in multiprocessing

I have a dummy example, I want to apply multiprocessing in it. Consider a scenario where you have a stream of numbers(which I call frame) incoming one by one. And I want to assign it to any single process that is available currently. So I am creating 4 processes that are running a while loop, seeing if any element in queue, than apply function on it.
The problem is that when I join it, it gets stuck in any while loop, even though I close the while loop before it. But somehow it gets stuck inside it.
Code:
# step 1, 4 processes
import multiprocessing as mp
import os
import time
class MpListOperations:
def __init__(self):
self.results_queue = mp.Manager().Queue()
self.frames_queue = mp.Manager().Queue()
self.flag = mp.Manager().Value(typecode='b',value=True)
self.list_nums = list(range(0,5000))
def process_list(self):
print(f"Process id {os.getpid()} started")
while self.flag.value:
# print(self.flag.value)
if self.frames_queue.qsize():
self.results_queue.put(self.frames_queue.get()**2)
def create_processes(self, no_of_processes = mp.cpu_count()):
print("Creating Processes")
self.processes = [mp.Process(target=self.process_list) for _ in range(no_of_processes)]
def start_processes(self):
print(f"starting processes")
for process in self.processes:
process.start()
def join_process(self):
print("Joining Processes")
while True:
if not self.frames_queue.qsize():
self.flag.value=False
print("JOININNG HERE")
for process in self.processes:
exit_code = process.join()
print(exit_code)
print("BREAKING DONE")
break
def stream_frames(self):
print("Streaming Frames")
for frame in self.list_nums:
self.frames_queue.put(frame)
if __name__=="__main__":
start = time.time()
mp_ops = MpListOperations()
mp_ops.create_processes()
mp_ops.start_processes()
mp_ops.stream_frames()
mp_ops.join_process()
print(time.time()-start)
Now if I add a timeout parameter in join, even 0, i.e exit_code = process.join(0) it works. I want to understand in this scenario, if this code is correct, what should be the value of timeout? Why is it working with timeout and not without it? What is the proper way to implement multiprocessing with it?

If you look at the documentation for a managed queue you will see that the qsize method only returns an approximate size. I would therefore not use it for testing when all the items have been taken of the frames queue. Presumably you want to let the processes run until all frames have been processed. The simplest way I know would be to put N sentinel items on the frames queue after the actual frames have been put where N is the number of processes getting from the queue. A sentinel item is a special value that cannot be mistaken for an actual frame and signals to the process that there are no more items for it to get from the queue (i.e. a quasi end-of-file item). In this case we can use None as the sentinel items. Each process then just continues to do get operations on the queue until it sees a sentinel item and then terminates. There is therefore no need for the self.flag attribute.
Here is the updated and simplified code. I have made some other minor changes that have been commented:
import multiprocessing as mp
import os
import time
class MpListOperations:
def __init__(self):
# Only create one manager process:
manager = mp.Manager()
self.results_queue = manager.Queue()
self.frames_queue = manager.Queue()
# No need to convert range to a list:
self.list_nums = range(0, 5000)
def process_list(self):
print(f"Process id {os.getpid()} started")
while True:
frame = self.frames_queue.get()
if frame is None: # Sentinel?
# Yes, we are done:
break
self.results_queue.put(frame ** 2)
def create_processes(self, no_of_processes = mp.cpu_count()):
print("Creating Processes")
self.no_of_processes = no_of_processes
self.processes = [mp.Process(target=self.process_list) for _ in range(no_of_processes)]
def start_processes(self):
print("Starting Processes")
for process in self.processes:
process.start()
def join_processes(self):
print("Joining Processes")
for process in self.processes:
# join returns None:
process.join()
def stream_frames(self):
print("Streaming Frames")
for frame in self.list_nums:
self.frames_queue.put(frame)
# Put sentinels:
for _ in range(self.no_of_processes):
self.frames_queue.put(None)
if __name__== "__main__":
start = time.time()
mp_ops = MpListOperations()
mp_ops.create_processes()
mp_ops.start_processes()
mp_ops.stream_frames()
mp_ops.join_processes()
print(time.time()-start)
Prints:
Creating Processes
Starting Processes
Process id 28 started
Process id 29 started
Streaming Frames
Process id 33 started
Process id 31 started
Process id 38 started
Process id 44 started
Process id 42 started
Process id 45 started
Joining Processes
2.3660173416137695
Note for Windows
I have modified method start_processes to temporarily set attribute self.processes to None:
def start_processes(self):
print("Starting Processes")
processes = self.processes
# Don't try to pickle list of processes:
self.processes = None
for process in processes:
process.start()
# Restore attribute:
self.processes = processes
Otherwise under Windows we get a pickle error trying to serialize/deserialize a list of processes containing two or more multiprocessing.Process instances. The error is "TypeError: cannot pickle 'weakref' object." This can be demonstrated with the following code where we first try to pickle a list of 1 process and then a list of 2 processes:
import multiprocessing as mp
import os
class Foo:
def __init__(self, number_of_processes):
self.processes = [mp.Process(target=self.worker) for _ in range(number_of_processes)]
self.start_processes()
self.join_processes()
def start_processes(self):
processes = self.processes
for process in self.processes:
process.start()
def join_processes(self):
for process in self.processes:
process.join()
def worker(self):
print(f"Process id {os.getpid()} started")
print(f"Process id {os.getpid()} ended")
if __name__== "__main__":
foo = Foo(1)
foo = Foo(2)
Prints:
Process id 7540 started
Process id 7540 ended
Traceback (most recent call last):
File "C:\Booboo\test\test.py", line 26, in <module>
foo = Foo(2)
File "C:\Booboo\test\test.py", line 7, in __init__
self.start_processes()
File "C:\Booboo\test\test.py", line 13, in start_processes
process.start()
File "C:\Program Files\Python38\lib\multiprocessing\process.py", line 121, in start
self._popen = self._Popen(self)
File "C:\Program Files\Python38\lib\multiprocessing\context.py", line 224, in _Popen
return _default_context.get_context().Process._Popen(process_obj)
File "C:\Program Files\Python38\lib\multiprocessing\context.py", line 327, in _Popen
return Popen(process_obj)
File "C:\Program Files\Python38\lib\multiprocessing\popen_spawn_win32.py", line 93, in __init__
reduction.dump(process_obj, to_child)
File "C:\Program Files\Python38\lib\multiprocessing\reduction.py", line 60, in dump
ForkingPickler(file, protocol).dump(obj)
TypeError: cannot pickle 'weakref' object
Process id 18152 started
Process id 18152 ended
Traceback (most recent call last):
File "<string>", line 1, in <module>
File "C:\Program Files\Python38\lib\multiprocessing\spawn.py", line 116, in spawn_main
exitcode = _main(fd, parent_sentinel)
File "C:\Program Files\Python38\lib\multiprocessing\spawn.py", line 126, in _main
self = reduction.pickle.load(from_parent)
EOFError: Ran out of input

The target loop is stuck in the get() method of your loop. This is because multiple processes could see that the queue wasn't empty, but only 1 of them was able to get the last item. The remaining processes are waiting for the next item to be available from the queue.
You might need to add a Lock when you are reading the size of the Queue object And getting the object of that queue.
Or alternatively, you avoid reading the size of the queue by simply using the queue.get() method with a timeout that allows us to check the flag regularly
import queue
TIMEOUT = 1 # seconds
class MpListOperations:
#[...]
def process_list(self):
print(f"Process id {os.getpid()} started")
previous = self.flag.value
while self.flag.value:
try:
got = self.frames_queue.get(timeout=TIMEOUT)
except queue.Empty:
pass
else:
print(f"Gotten {got}")
self.results_queue.put(got**2)
_next = self.flag.value
if previous != _next:
print(f"Flag change: {_next}")
$ python ./test_mp.py
Creating Processes
starting processes
Process id 36566 started
Streaming Frames
Process id 36565 started
Process id 36564 started
Process id 36570 started
Process id 36567 started
Gotten 0
Process id 36572 started
Gotten 1
Gotten 2
Gotten 3
Process id 36579 started
Gotten 4
Gotten 5
Gotten 6
Process id 36583 started
Gotten 7
# [...]
Gotten 4997
Joining Processes
Gotten 4998
Gotten 4999
JOININNG HERE
Flag change: False
Flag change: False
Flag change: False
Flag change: False
Flag change: False
Flag change: False
Flag change: False
Flag change: False
Exit code : None
Exit code : None
Exit code : None
Exit code : None
Exit code : None
Exit code : None
Exit code : None
Exit code : None
BREAKING DONE
1.4375360012054443
Alternatively, using a multiprocessing.Pool object:
def my_func(arg):
time.sleep(0.002)
return arg**2
def get_input():
for i in range(5000):
yield i
time.sleep(0.001)
if __name__=="__main__":
start = time.time()
mp_pool = mp.Pool()
result = mp_pool.map(my_func, get_input())
mp_pool.close()
mp_pool.join()
print(len(result))
print(f"Duration: {time.time()-start}")
Giving:
$ python ./test_mp.py
5000
Duration: 6.847279787063599

Processes not joining even after flushing their queues

I wrote a program using the module multiprocessing which globally executes as follows:
both a simulation and an ui processes are started.
the simulation process feeds a queue with new simulation states. If the queue is full, the simulation loop isn't blocked so it can handle possible incoming messages.
the ui process consumes the simulation queue.
after around 1 second of execution time, the ui process sends a quit event to the main process then exits the loop. Upon exiting it sends a stopped event to the main process through the _create_process()'s inner wrapper() function.
the main process receives both events in whichever order. The quit event results in the main process sending stop signals to all the child processes, while the stopped event increments a counter in the main loop which will cause it to exit after having received as many stopped events as there are processes.
the simulation process receives the stop event and exits the loop, sending in turn a stopped event to the main process.
the main process has now received 2 stopped events in total and concludes that all child processes are—on their way to be—stopped. As a result, the main loop is exited
the run() function flushes the queues which have been written by the child processes.
the child processes are being joined.
The problem is that quite often (but not always) the program will hang upon trying to join the simulation process, as per the log below.
[...]
[INFO/ui] process exiting with exitcode 0
[DEBUG/MainProcess] starting thread to feed data to pipe
[DEBUG/MainProcess] ... done self._thread.start()
[DEBUG/simulation] Queue._start_thread()
[DEBUG/simulation] doing self._thread.start()
[DEBUG/simulation] starting thread to feed data to pipe
[DEBUG/simulation] ... done self._thread.start()
[DEBUG/simulation] telling queue thread to quit
[DEBUG/MainProcess] all child processes (2) should have been stopped!
[INFO/simulation] process shutting down
[DEBUG/simulation] running all "atexit" finalizers with priority >= 0
[DEBUG/simulation] telling queue thread to quit
[DEBUG/simulation] running the remaining "atexit" finalizers
[DEBUG/simulation] joining queue thread
[DEBUG/MainProcess] joining process <Process(simulation, started)>
[DEBUG/simulation] feeder thread got sentinel -- exiting
[DEBUG/simulation] ... queue thread joined
[DEBUG/simulation] joining queue thread
Stopping the execution through a Ctrl + C in the shell results in these mangled tracebacks:
Process simulation:
Traceback (most recent call last):
Traceback (most recent call last):
File "./debug.py", line 224, in <module>
run()
File "/Library/Frameworks/Python.framework/Versions/3.4/lib/python3.4/multiprocessing/process.py", line 257, in _bootstrap
util._exit_function()
File "./debug.py", line 92, in run
process.join() #< This doesn't work.
File "/Library/Frameworks/Python.framework/Versions/3.4/lib/python3.4/multiprocessing/util.py", line 312, in _exit_function
_run_finalizers()
File "/Library/Frameworks/Python.framework/Versions/3.4/lib/python3.4/multiprocessing/process.py", line 121, in join
File "/Library/Frameworks/Python.framework/Versions/3.4/lib/python3.4/multiprocessing/util.py", line 252, in _run_finalizers
finalizer()
File "/Library/Frameworks/Python.framework/Versions/3.4/lib/python3.4/multiprocessing/util.py", line 185, in __call__
res = self._callback(*self._args, **self._kwargs)
res = self._popen.wait(timeout)
File "/Library/Frameworks/Python.framework/Versions/3.4/lib/python3.4/multiprocessing/popen_fork.py", line 54, in wait
File "/Library/Frameworks/Python.framework/Versions/3.4/lib/python3.4/multiprocessing/queues.py", line 196, in _finalize_join
thread.join()
return self.poll(os.WNOHANG if timeout == 0.0 else 0)
File "/Library/Frameworks/Python.framework/Versions/3.4/lib/python3.4/multiprocessing/popen_fork.py", line 30, in poll
pid, sts = os.waitpid(self.pid, flag)
KeyboardInterrupt
File "/Library/Frameworks/Python.framework/Versions/3.4/lib/python3.4/threading.py", line 1060, in join
self._wait_for_tstate_lock()
File "/Library/Frameworks/Python.framework/Versions/3.4/lib/python3.4/threading.py", line 1076, in _wait_for_tstate_lock
elif lock.acquire(block, timeout):
KeyboardInterrupt
As for the code, here is a stripped down version of it (hence why it often seems incomplete):
#!/usr/bin/env python3
import logging
import multiprocessing
import pickle
import queue
import time
from collections import namedtuple
_LOGGER = multiprocessing.log_to_stderr()
_LOGGER.setLevel(logging.DEBUG)
_BUFFER_SIZE = 4
_DATA_LENGTH = 2 ** 12
_STATUS_SUCCESS = 0
_STATUS_FAILURE = 1
_EVENT_ERROR = 0
_EVENT_QUIT = 1
_EVENT_STOPPED = 2
_MESSAGE_STOP = 0
_MESSAGE_EVENT = 1
_MESSAGE_SIMULATION_UPDATE = 2
_Message = namedtuple('_Message', ('type', 'value',))
_StopMessage = namedtuple('_StopMessage', ())
_EventMessage = namedtuple('_EventMessage', ('type', 'value',))
_SimulationUpdateMessage = namedtuple('_SimulationUpdateMessage', ('state',))
_MESSAGE_STRUCTS = {
_MESSAGE_STOP: _StopMessage,
_MESSAGE_EVENT: _EventMessage,
_MESSAGE_SIMULATION_UPDATE: _SimulationUpdateMessage
}
def run():
# Messages from the main process to the child ones.
downward_queue = multiprocessing.Queue()
# Messages from the child processes to the main one.
upward_queue = multiprocessing.Queue()
# Messages from the simulation process to the UI one.
simulation_to_ui_queue = multiprocessing.Queue(maxsize=_BUFFER_SIZE)
# Regroup all the queues that can be written by child processes.
child_process_queues = (upward_queue, simulation_to_ui_queue,)
processes = (
_create_process(
_simulation,
upward_queue,
name='simulation',
args=(
simulation_to_ui_queue,
downward_queue
)
),
_create_process(
_ui,
upward_queue,
name='ui',
args=(
upward_queue,
simulation_to_ui_queue,
downward_queue
)
)
)
try:
for process in processes:
process.start()
_main(downward_queue, upward_queue, len(processes))
finally:
# while True:
# alive_processes = tuple(process for process in processes
# if process.is_alive())
# if not alive_processes:
# break
# _LOGGER.debug("processes still alive: %s" % (alive_processes,))
for q in child_process_queues:
_flush_queue(q)
for process in processes:
_LOGGER.debug("joining process %s" % process)
# process.terminate() #< This works!
process.join() #< This doesn't work.
def _main(downward_queue, upward_queue, process_count):
try:
stopped_count = 0
while True:
message = _receive_message(upward_queue, False)
if message is not None and message.type == _MESSAGE_EVENT:
event_type = message.value.type
if event_type in (_EVENT_QUIT, _EVENT_ERROR):
break
elif event_type == _EVENT_STOPPED:
stopped_count += 1
if stopped_count >= process_count:
break
finally:
# Whatever happens, make sure that all child processes have stopped.
if stopped_count >= process_count:
return
# Send a 'stop' signal to all the child processes.
for _ in range(process_count):
_send_message(downward_queue, True, _MESSAGE_STOP)
while True:
message = _receive_message(upward_queue, False)
if (message is not None
and message.type == _MESSAGE_EVENT
and message.value.type == _EVENT_STOPPED):
stopped_count += 1
if stopped_count >= process_count:
_LOGGER.debug(
"all child processes (%d) should have been stopped!"
% stopped_count
)
break
def _simulation(simulation_to_ui_queue, downward_queue):
simulation_state = [i * 0.123 for i in range(_DATA_LENGTH)]
# When the queue is full (possibly form reaching _BUFFER_SIZE), the next
# solve is computed and kept around until the queue is being consumed.
next_solve_message = None
while True:
message = _receive_message(downward_queue, False)
if message is not None and message.type == _MESSAGE_STOP:
break
if next_solve_message is None:
# _step(simulation_state)
# Somehow the copy (pickle) seems to increase the chances for
# the issue to happen.
next_solve_message = _SimulationUpdateMessage(
state=pickle.dumps(simulation_state)
)
status = _send_message(simulation_to_ui_queue, False,
_MESSAGE_SIMULATION_UPDATE,
**next_solve_message._asdict())
if status == _STATUS_SUCCESS:
next_solve_message = None
def _ui(upward_queue, simulation_to_ui_queue, downward_queue):
time_start = -1.0
previous_time = 0.0
while True:
message = _receive_message(downward_queue, False)
if message is not None and message.type == _MESSAGE_STOP:
break
if time_start < 0:
current_time = 0.0
time_start = time.perf_counter()
else:
current_time = time.perf_counter() - time_start
message = _receive_message(simulation_to_ui_queue, False)
if current_time > 1.0:
_LOGGER.debug("asking to quit")
_send_message(upward_queue, True, _MESSAGE_EVENT,
type=_EVENT_QUIT, value=None)
break
previous_time = current_time
def _create_process(target, upward_queue, name='', args=None):
def wrapper(function, upward_queue, *args, **kwargs):
try:
function(*args, **kwargs)
except Exception:
_send_message(upward_queue, True, _MESSAGE_EVENT,
type=_EVENT_ERROR, value=None)
finally:
_send_message(upward_queue, True, _MESSAGE_EVENT,
type=_EVENT_STOPPED, value=None)
upward_queue.close()
process = multiprocessing.Process(
target=wrapper,
name=name,
args=(target, upward_queue) + args,
kwargs={}
)
return process
def _receive_message(q, block):
try:
message = q.get(block=block)
except queue.Empty:
return None
return message
def _send_message(q, block, message_type, **kwargs):
message_value = _MESSAGE_STRUCTS[message_type](**kwargs)
try:
q.put(_Message(type=message_type, value=message_value), block=block)
except queue.Full:
return _STATUS_FAILURE
return _STATUS_SUCCESS
def _flush_queue(q):
try:
while True:
q.get(block=False)
except queue.Empty:
pass
if __name__ == '__main__':
run()
Related questions on StackOverflow and hints in Python's doc basically boil down to needing to flush the queues before joining the processes, which I believe I've been trying to do here. I realize that the simulation queue could still be trying to push the (potentially large) buffered data onto the pipe by the time the program would try to flush them upon exiting, and thus ending up with still non-empty queues. This is why I tried to ensure that all the child processes were stopped before reaching this point. Now, looking at the log above and at the additional log outputted after uncommenting the while True loop checking for alive processes, it appears that the simulation process simply doesn't want to completely shut down even though its target function definitely exited. Could this be the reason of my problem?
If so, how am I suppsoed to deal with it cleanly? Otherwise, what am I missing here?
Tested with Python 3.4 on Mac OS X 10.9.5.
PS: I'm wondering if this couldn't be related to this bug ?

Sounds like the issue was indeed due to some delay in pushing the data through the queue, causing the flushes to be ineffective because fired too early.
A simple while process.is_alive(): flush_the_queues() seems to do the trick!

Lately I have run into a similar use case like yours: multiple processes (up to 11), one input queue, one output queue. But very heavy output queue.
I was getting an overhead of up to 5 seconds (!) using your suggestion to perform while process.is_alive(): flush_the_queues() before the process.join().
I've reduced that overhead down to 0.7 seconds by relying on a multiprocessing.Manager.list instead of a multiprocessing.Queue for the output queue. The multiprocessing.Manager.list doesn't need any flushing. I might consider also finding an alternative to the input queue if I can..
Full example here:
import multiprocessing
import queue
import time
PROCESSES = multiprocessing.cpu_count() - 1
processes = []
def run():
start = time.time()
input_queue = multiprocessing.Queue()
feed_input_queue(input_queue)
with multiprocessing.Manager() as manager:
output_list = manager.list()
for _ in range(PROCESSES):
p = multiprocessing.Process(target=_execute, args=(input_queue, output_list))
processes.append(p)
p.start()
print(f"Time to process = {time.time() - start:.10f}")
start = time.time()
for p in processes:
while p.is_alive(): # in principle we could get rid of this if we find an alternative to the output queue
_flush_queue(input_queue)
p.join()
print(f"Time to join = {time.time() - start:.10f}")
# from here you can do something with the output_list
def _feed_input_queue(input_queue):
for i in range(10000):
input_queue.put(i)
def _execute(input_queue: multiprocessing.Queue, output_list: list):
while not input_queue.empty():
input_item = input_queue.get()
output_list.append(do_and_return_something_heavy(input_item))
return True
def _flush_queue(q):
try:
while True:
q.get(block=False)
except queue.Empty:
pass
def do_and_return_something_heavy(input_item):
return str(input_item) * 100000
if __name__ == '__main__':
run()
Output
Time to process = 0.1855618954
Time to join = 0.6889970303
Tested on Python 3.6.

outside vs inside __main__ variable definition in multiprocessing

I have following code:
import multiprocessing
import time
import os
# WHEN SEMAPHORE IS DEFINED HERE THEN IT IT WORKS
semaphore = multiprocessing.Semaphore(1)
def producer(num, output):
semaphore.acquire()
time.sleep(1)
element = "PROCESS: %d PID: %d PPID: %d" % (num, os.getpid(), os.getppid())
print "WRITE -> " + element
output.put(element)
time.sleep(1)
semaphore.release()
if __name__ == '__main__':
"""
Reads elements as soon as they are are put inside queue
"""
output = multiprocessing.Manager().Queue()
pool = multiprocessing.Pool(4)
lst = range(40)
# WHEN SEMAPHORE IS DEFINED HERE THEN IT DOES NOT WORKS
# semaphore = multiprocessing.Semaphore(1)
for i in lst:
pool.apply_async(producer, (i, output))
# print "%d Do not wait!" % i
# res.get()
counter = 0
while True:
try:
print "READ <- " + output.get_nowait()
counter += 1
if (counter == len(lst)):
print "Break"
break
except:
print "READ <- NOTHING IN BUFFER"
pass
time.sleep(1)
This code is working as expected and it prints:
READ <- NOTHING IN BUFFER
WRITE -> PROCESS: 0 PID: 15803 PPID: 15798
READ <- NOTHING IN BUFFER
READ <- PROCESS: 0 PID: 15803 PPID: 15798
READ <- NOTHING IN BUFFER
WRITE -> PROCESS: 1 PID: 15806 PPID: 15798
READ <- PROCESS: 1 PID: 15806 PPID: 15798
...
Then I have this version which is not working (It is basically the same as first one except the definition of semaphore is in another place):
import multiprocessing
import time
import os
# WHEN SEMAPHORE IS DEFINED HERE THEN IT IT WORKS
# semaphore = multiprocessing.Semaphore(1)
def producer(num, output):
print hex(id(semaphore))
semaphore.acquire()
time.sleep(1)
element = "PROCESS: %d PID: %d PPID: %d" % (num, os.getpid(), os.getppid())
print "WRITE -> " + element
output.put(element)
time.sleep(1)
semaphore.release()
if __name__ == '__main__':
"""
Reads elements as soon as they are are put inside queue
"""
output = multiprocessing.Manager().Queue()
pool = multiprocessing.Pool(4)
lst = range(40)
# WHEN SEMAPHORE IS DEFINED HERE THEN IT DOES NOT WORKS
semaphore = multiprocessing.Semaphore(1)
for i in lst:
pool.apply_async(producer, (i, output))
# print "%d Do not wait!" % i
# res.get()
counter = 0
while True:
try:
print "READ <- " + output.get_nowait()
counter += 1
if (counter == len(lst)):
print "Break"
break
except:
print "READ <- NOTHING IN BUFFER"
pass
time.sleep(1)
This version prints:
READ <- NOTHING IN BUFFER
READ <- NOTHING IN BUFFER
READ <- NOTHING IN BUFFER
READ <- NOTHING IN BUFFER
READ <- NOTHING IN BUFFER
READ <- NOTHING IN BUFFER
READ <- NOTHING IN BUFFER
...
It seems like if producer never writes anything to Queue. I've read somewhere that apply_sync does not print error messages. So I've changed pool.apply_async(producer, (i, output)) to pool.apply(producer, (i, output)) in second code, to see what is going on. It seems that semaphore is not defined, here is the output:
Traceback (most recent call last):
File "glob_var_wrong.py", line 31, in <module>
pool.apply(producer, (i, output))
File "/usr/lib/python2.7/multiprocessing/pool.py", line 244, in apply
return self.apply_async(func, args, kwds).get()
File "/usr/lib/python2.7/multiprocessing/pool.py", line 567, in get
raise self._value
NameError: global name 'semaphore' is not defined
However following code runs correctly and print 10 (value defined inside __main__):
global_var = 20
def print_global_var():
print global_var
if __name__ == '__main__':
global_var = 10
print_global_var()
It seems that in this code global variable can be defined inside __main__ while in previous codes it is not possible. First I was assuming that variables defined inside __main__ are not shared between processes but it only affects semaphore and not output, pool, lst. Why is this happening?

When you create a new process using Multiprocessing.Process (used underneath the hood by Pool, it copies the local scope, pickles it, and sends it to a new process to evaluate.
Because you did not define the variable semaphore before calling Pool(4), the variable is undefined (in those OTHER processes where the code gets evaluated) and the function producer will throw an exception.
To see this, change the definition
def producer(num, output):
print hex(id(semaphore))
try:
semaphore.acquire()
except Exception as e:
print e
time.sleep(1)
element = "PROCESS: %d PID: %d PPID: %d" % (num, os.getpid(), os.getppid())
print "WRITE -> " + element
output.put(element)
time.sleep(1)
semaphore.release()
and now your failing code will print out a bunch (40) of errors that looks like
global name 'semaphore' is not defined
This is why semaphore has to be defined BEFORE calling Pool

It is because you execute the code on Windows.
You will get the expected results on Linux.
That's the difference between fork and spawn.

Python Multiprocess issues process not starting

I wrote this code, I want to have a main thread that starts multiple subprocesses that spawn a listener thread to wait for a kill message subprocess works but testprocess does not run there are no errors any ideas?
from multiprocessing import Process, Pipe
from threading import Thread
import time
Alive = True
def listener_thread(conn): #listens for kill from main
global Alive
while True:
data = conn.recv()
if data == "kill":
Alive = False #value for kill
break
def subprocess(conn):
t = Thread(target=listener_thread, args=(conn,))
count = 0
t.start()
while Alive:
print "Run number = %d" % count
count = count + 1
def testprocess(conn):
t = Thread(target=listner_thread, args=(conn,))
count = 0
t.start()
while Alive:
print "This is a different thread run = %d" % count
count = count + 1
parent_conn, child_conn = Pipe()
p = Process(target=subprocess, args=(child_conn,))
p2 = Process(target=testprocess, args=(child_conn,))
runNum = int(raw_input("Enter a number: "))
p.start()
p2.start()
time.sleep(runNum)
parent_conn.send("kill") #sends kill to listener thread to tell them when to stop
p.join()
p2.join()

A typo in testprocess makes the function to quit early.
listner_thread should be listener_thread.
If you comment out subprocess related code and run the code, you will see following error:
Process Process-1:
Traceback (most recent call last):
File "/usr/lib/python2.7/multiprocessing/process.py", line 258, in _bootstrap
self.run()
File "/usr/lib/python2.7/multiprocessing/process.py", line 114, in run
self._target(*self._args, **self._kwargs)
File "t.py", line 25, in testprocess
t = Thread(target=listner_thread, args=(conn,))
NameError: global name 'listner_thread' is not defined

cant reference a global dictionary entry

I am aware that global variables are not always the best way to deal with things in this case they are fine for what I am doing. I am not going to be doing heavy read/writes mostly just reads.
alive = {'subAlive': True, 'testAlive': True};
def sub_listener(conn): #listens for kill from main
global alive
while True:
data = conn.recv()
if data == "kill":
alive['subAlive'] = False; #value for kill
break
def subprocess(conn, threadNum):
t = Thread(target=sub_listener, args=(conn,))
count = 0
threadVal = threadNum
t.start()
run = alive[subAlive];
while run:
print "Thread %d Run number = %d" % (threadVal, count)
count = count + 1
sub_parent, sub_child = Pipe()
runNum = int(raw_input("Enter a number: "))
threadNum = int(raw_input("Enter number of threads: "))
print "Starting threads"
for i in range(threadNum):
p = Process(target=subprocess, args=(sub_child, i))
p.start()
print "Starting run"
time.sleep(runNum)
print "Terminating Subprocess run"
for i in range(threadNum):
sub_parent.send("kill") #sends kill to listener
p.join()
I get this error
NameError: global name 'testAlive' is not defined
Traceback (most recent call last):
File "/usr/lib/python2.7/multiprocessing/process.py", line 258, in _bootstrap
self.run()
File "/usr/lib/python2.7/multiprocessing/process.py", line 114, in run
self._target(*self._args, **self._kwargs)
File "multiprocessDemo.py", line 38, in subprocess
run = alive[subAlive];
NameError: global name 'subAlive' is not defined
I have tried accessing the dictionary a few different ways and I can't seem to find out what is wrong on google. If I use separate variables it does work but that wont dynamically scale well.

Put quotes around subAlive:
run = alive['subAlive']