bpo-31540: Allow passing multiprocessing context to ProcessPoolExecutor (#3682)

.travis.yml

@@ -56,7 +56,7 @@ matrix:
             ./venv/bin/python -m test.pythoninfo
       script:
         # Skip tests that re-run the entire test suite.
-        - ./venv/bin/python -m coverage run --pylib -m test --fail-env-changed -uall,-cpu -x test_multiprocessing_fork -x test_multiprocessing_forkserver -x test_multiprocessing_spawn
+        - ./venv/bin/python -m coverage run --pylib -m test --fail-env-changed -uall,-cpu -x test_multiprocessing_fork -x test_multiprocessing_forkserver -x test_multiprocessing_spawn -x test_concurrent_futures
       after_script:  # Probably should be after_success once test suite updated to run under coverage.py.
         # Make the `coverage` command available to Codecov w/ a version of Python that can parse all source files.
         - source ./venv/bin/activate

Doc/library/concurrent.futures.rst

@@ -191,13 +191,16 @@ that :class:`ProcessPoolExecutor` will not work in the interactive interpreter.
 Calling :class:`Executor` or :class:`Future` methods from a callable submitted
 to a :class:`ProcessPoolExecutor` will result in deadlock.
 
-.. class:: ProcessPoolExecutor(max_workers=None)
+.. class:: ProcessPoolExecutor(max_workers=None, mp_context=None)
 
    An :class:`Executor` subclass that executes calls asynchronously using a pool
    of at most *max_workers* processes.  If *max_workers* is ``None`` or not
    given, it will default to the number of processors on the machine.
    If *max_workers* is lower or equal to ``0``, then a :exc:`ValueError`
    will be raised.
+   *mp_context* can be a multiprocessing context or None. It will be used to
+   launch the workers. If *mp_context* is ``None`` or not given, the default
+   multiprocessing context is used.
 
    .. versionchanged:: 3.3
       When one of the worker processes terminates abruptly, a
@@ -205,6 +208,10 @@ to a :class:`ProcessPoolExecutor` will result in deadlock.
       was undefined but operations on the executor or its futures would often
       freeze or deadlock.
 
+   .. versionchanged:: 3.7
+      The *mp_context* argument was added to allow users to control the
+      start_method for worker processes created by the pool.
+
 
 .. _processpoolexecutor-example:
 

Lib/concurrent/futures/process.py

@@ -50,8 +50,7 @@ import os
 from concurrent.futures import _base
 import queue
 from queue import Full
-import multiprocessing
-from multiprocessing import SimpleQueue
+import multiprocessing as mp
 from multiprocessing.connection import wait
 import threading
 import weakref
@@ -74,11 +73,11 @@ import traceback
 # threads/processes finish.
 
 _threads_queues = weakref.WeakKeyDictionary()
-_shutdown = False
+_global_shutdown = False
 
 def _python_exit():
-    global _shutdown
-    _shutdown = True
+    global _global_shutdown
+    _global_shutdown = True
     items = list(_threads_queues.items())
     for t, q in items:
         q.put(None)
@@ -158,12 +157,10 @@ def _process_worker(call_queue, result_queue):
     This worker is run in a separate process.
 
     Args:
-        call_queue: A multiprocessing.Queue of _CallItems that will be read and
+        call_queue: A ctx.Queue of _CallItems that will be read and
             evaluated by the worker.
-        result_queue: A multiprocessing.Queue of _ResultItems that will written
+        result_queue: A ctx.Queue of _ResultItems that will written
             to by the worker.
-        shutdown: A multiprocessing.Event that will be set as a signal to the
-            worker that it should exit when call_queue is empty.
     """
     while True:
         call_item = call_queue.get(block=True)
@@ -180,6 +177,11 @@ def _process_worker(call_queue, result_queue):
             result_queue.put(_ResultItem(call_item.work_id,
                                          result=r))
 
+        # Liberate the resource as soon as possible, to avoid holding onto
+        # open files or shared memory that is not needed anymore
+        del call_item
+
+
 def _add_call_item_to_queue(pending_work_items,
                             work_ids,
                             call_queue):
@@ -231,20 +233,21 @@ def _queue_management_worker(executor_reference,
         executor_reference: A weakref.ref to the ProcessPoolExecutor that owns
             this thread. Used to determine if the ProcessPoolExecutor has been
             garbage collected and that this function can exit.
-        process: A list of the multiprocessing.Process instances used as
+        process: A list of the ctx.Process instances used as
             workers.
         pending_work_items: A dict mapping work ids to _WorkItems e.g.
             {5: <_WorkItem...>, 6: <_WorkItem...>, ...}
         work_ids_queue: A queue.Queue of work ids e.g. Queue([5, 6, ...]).
-        call_queue: A multiprocessing.Queue that will be filled with _CallItems
+        call_queue: A ctx.Queue that will be filled with _CallItems
             derived from _WorkItems for processing by the process workers.
-        result_queue: A multiprocessing.Queue of _ResultItems generated by the
+        result_queue: A ctx.SimpleQueue of _ResultItems generated by the
             process workers.
     """
     executor = None
 
     def shutting_down():
-        return _shutdown or executor is None or executor._shutdown_thread
+        return (_global_shutdown or executor is None
+                or executor._shutdown_thread)
 
     def shutdown_worker():
         # This is an upper bound
@@ -254,7 +257,7 @@ def _queue_management_worker(executor_reference,
         # Release the queue's resources as soon as possible.
         call_queue.close()
         # If .join() is not called on the created processes then
-        # some multiprocessing.Queue methods may deadlock on Mac OS X.
+        # some ctx.Queue methods may deadlock on Mac OS X.
         for p in processes.values():
             p.join()
 
@@ -377,13 +380,15 @@ class BrokenProcessPool(RuntimeError):
 
 
 class ProcessPoolExecutor(_base.Executor):
-    def __init__(self, max_workers=None):
+    def __init__(self, max_workers=None, mp_context=None):
         """Initializes a new ProcessPoolExecutor instance.
 
         Args:
             max_workers: The maximum number of processes that can be used to
                 execute the given calls. If None or not given then as many
                 worker processes will be created as the machine has processors.
+            mp_context: A multiprocessing context to launch the workers. This
+                object should provide SimpleQueue, Queue and Process.
         """
         _check_system_limits()
 
@@ -394,17 +399,20 @@ class ProcessPoolExecutor(_base.Executor):
                 raise ValueError("max_workers must be greater than 0")
 
             self._max_workers = max_workers
+        if mp_context is None:
+            mp_context = mp.get_context()
+        self._mp_context = mp_context
 
         # Make the call queue slightly larger than the number of processes to
         # prevent the worker processes from idling. But don't make it too big
         # because futures in the call queue cannot be cancelled.
-        self._call_queue = multiprocessing.Queue(self._max_workers +
-                                                 EXTRA_QUEUED_CALLS)
+        queue_size = self._max_workers + EXTRA_QUEUED_CALLS
+        self._call_queue = mp_context.Queue(queue_size)
         # Killed worker processes can produce spurious "broken pipe"
         # tracebacks in the queue's own worker thread. But we detect killed
         # processes anyway, so silence the tracebacks.
         self._call_queue._ignore_epipe = True
-        self._result_queue = SimpleQueue()
+        self._result_queue = mp_context.SimpleQueue()
         self._work_ids = queue.Queue()
         self._queue_management_thread = None
         # Map of pids to processes
@@ -439,7 +447,7 @@ class ProcessPoolExecutor(_base.Executor):
 
     def _adjust_process_count(self):
         for _ in range(len(self._processes), self._max_workers):
-            p = multiprocessing.Process(
+            p = self._mp_context.Process(
                 target=_process_worker,
                 args=(self._call_queue,
                       self._result_queue))

Lib/test/test_concurrent_futures.py

@@ -19,6 +19,7 @@ from concurrent import futures
 from concurrent.futures._base import (
     PENDING, RUNNING, CANCELLED, CANCELLED_AND_NOTIFIED, FINISHED, Future)
 from concurrent.futures.process import BrokenProcessPool
+from multiprocessing import get_context
 
 
 def create_future(state=PENDING, exception=None, result=None):
@@ -56,6 +57,15 @@ class MyObject(object):
         pass
 
 
+class EventfulGCObj():
+    def __init__(self, ctx):
+        mgr = get_context(ctx).Manager()
+        self.event = mgr.Event()
+
+    def __del__(self):
+        self.event.set()
+
+
 def make_dummy_object(_):
     return MyObject()
 
@@ -77,7 +87,13 @@ class ExecutorMixin:
 
         self.t1 = time.time()
         try:
-            self.executor = self.executor_type(max_workers=self.worker_count)
+            if hasattr(self, "ctx"):
+                self.executor = self.executor_type(
+                    max_workers=self.worker_count,
+                    mp_context=get_context(self.ctx))
+            else:
+                self.executor = self.executor_type(
+                    max_workers=self.worker_count)
         except NotImplementedError as e:
             self.skipTest(str(e))
         self._prime_executor()
@@ -107,8 +123,29 @@ class ThreadPoolMixin(ExecutorMixin):
     executor_type = futures.ThreadPoolExecutor
 
 
-class ProcessPoolMixin(ExecutorMixin):
+class ProcessPoolForkMixin(ExecutorMixin):
+    executor_type = futures.ProcessPoolExecutor
+    ctx = "fork"
+
+    def setUp(self):
+        if sys.platform == "win32":
+            self.skipTest("require unix system")
+        super().setUp()
+
+
+class ProcessPoolSpawnMixin(ExecutorMixin):
+    executor_type = futures.ProcessPoolExecutor
+    ctx = "spawn"
+
+
+class ProcessPoolForkserverMixin(ExecutorMixin):
     executor_type = futures.ProcessPoolExecutor
+    ctx = "forkserver"
+
+    def setUp(self):
+        if sys.platform == "win32":
+            self.skipTest("require unix system")
+        super().setUp()
 
 
 class ExecutorShutdownTest:
@@ -124,9 +161,17 @@ class ExecutorShutdownTest:
             from concurrent.futures import {executor_type}
             from time import sleep
             from test.test_concurrent_futures import sleep_and_print
+            if __name__ == "__main__":
+                context = '{context}'
+                if context == "":
                     t = {executor_type}(5)
+                else:
+                    from multiprocessing import get_context
+                    context = get_context(context)
+                    t = {executor_type}(5, mp_context=context)
                 t.submit(sleep_and_print, 1.0, "apple")
-            """.format(executor_type=self.executor_type.__name__))
+            """.format(executor_type=self.executor_type.__name__,
+                       context=getattr(self, "ctx", "")))
         # Errors in atexit hooks don't change the process exit code, check
         # stderr manually.
         self.assertFalse(err)
@@ -194,7 +239,7 @@ class ThreadPoolShutdownTest(ThreadPoolMixin, ExecutorShutdownTest, BaseTestCase
             t.join()
 
 
-class ProcessPoolShutdownTest(ProcessPoolMixin, ExecutorShutdownTest, BaseTestCase):
+class ProcessPoolShutdownTest(ExecutorShutdownTest):
     def _prime_executor(self):
         pass
 
@@ -233,6 +278,22 @@ class ProcessPoolShutdownTest(ProcessPoolMixin, ExecutorShutdownTest, BaseTestCa
         call_queue.join_thread()
 
 
+class ProcessPoolForkShutdownTest(ProcessPoolForkMixin, BaseTestCase,
+                                  ProcessPoolShutdownTest):
+    pass
+
+
+class ProcessPoolForkserverShutdownTest(ProcessPoolForkserverMixin,
+                                        BaseTestCase,
+                                        ProcessPoolShutdownTest):
+    pass
+
+
+class ProcessPoolSpawnShutdownTest(ProcessPoolSpawnMixin, BaseTestCase,
+                                   ProcessPoolShutdownTest):
+    pass
+
+
 class WaitTests:
 
     def test_first_completed(self):
@@ -352,7 +413,17 @@ class ThreadPoolWaitTests(ThreadPoolMixin, WaitTests, BaseTestCase):
             sys.setswitchinterval(oldswitchinterval)
 
 
-class ProcessPoolWaitTests(ProcessPoolMixin, WaitTests, BaseTestCase):
+class ProcessPoolForkWaitTests(ProcessPoolForkMixin, WaitTests, BaseTestCase):
+    pass
+
+
+class ProcessPoolForkserverWaitTests(ProcessPoolForkserverMixin, WaitTests,
+                                     BaseTestCase):
+    pass
+
+
+class ProcessPoolSpawnWaitTests(ProcessPoolSpawnMixin, BaseTestCase,
+                                WaitTests):
     pass
 
 
@@ -440,7 +511,19 @@ class ThreadPoolAsCompletedTests(ThreadPoolMixin, AsCompletedTests, BaseTestCase
     pass
 
 
-class ProcessPoolAsCompletedTests(ProcessPoolMixin, AsCompletedTests, BaseTestCase):
+class ProcessPoolForkAsCompletedTests(ProcessPoolForkMixin, AsCompletedTests,
+                                      BaseTestCase):
+    pass
+
+
+class ProcessPoolForkserverAsCompletedTests(ProcessPoolForkserverMixin,
+                                            AsCompletedTests,
+                                            BaseTestCase):
+    pass
+
+
+class ProcessPoolSpawnAsCompletedTests(ProcessPoolSpawnMixin, AsCompletedTests,
+                                       BaseTestCase):
     pass
 
 
@@ -540,7 +623,7 @@ class ThreadPoolExecutorTest(ThreadPoolMixin, ExecutorTest, BaseTestCase):
                          (os.cpu_count() or 1) * 5)
 
 
-class ProcessPoolExecutorTest(ProcessPoolMixin, ExecutorTest, BaseTestCase):
+class ProcessPoolExecutorTest(ExecutorTest):
     def test_killed_child(self):
         # When a child process is abruptly terminated, the whole pool gets
         # "broken".
@@ -595,6 +678,34 @@ class ProcessPoolExecutorTest(ProcessPoolMixin, ExecutorTest, BaseTestCase):
         self.assertIn('raise RuntimeError(123) # some comment',
                       f1.getvalue())
 
+    def test_ressources_gced_in_workers(self):
+        # Ensure that argument for a job are correctly gc-ed after the job
+        # is finished
+        obj = EventfulGCObj(self.ctx)
+        future = self.executor.submit(id, obj)
+        future.result()
+
+        self.assertTrue(obj.event.wait(timeout=1))
+
+
+class ProcessPoolForkExecutorTest(ProcessPoolForkMixin,
+                                  ProcessPoolExecutorTest,
+                                  BaseTestCase):
+    pass
+
+
+class ProcessPoolForkserverExecutorTest(ProcessPoolForkserverMixin,
+                                        ProcessPoolExecutorTest,
+                                        BaseTestCase):
+    pass
+
+
+class ProcessPoolSpawnExecutorTest(ProcessPoolSpawnMixin,
+                                   ProcessPoolExecutorTest,
+                                   BaseTestCase):
+    pass
+
+
 
 class FutureTests(BaseTestCase):
     def test_done_callback_with_result(self):

Misc/NEWS.d/next/Library/2017-09-22-16-02-00.bpo-31540.ybDHT5.rst

+Allow passing a context object in
+:class:`concurrent.futures.ProcessPoolExecutor` constructor.
+Also, free job ressources in :class:`concurrent.futures.ProcessPoolExecutor`
+earlier to improve memory usage when a worker waits for new jobs.