damccorm commented on code in PR #37112:
URL: https://github.com/apache/beam/pull/37112#discussion_r2747554004


##########
sdks/python/apache_beam/utils/multi_process_shared_test.py:
##########
@@ -243,6 +283,181 @@ def test_release_always_proxy(self):
       counter1.get()
 
 
+class MultiProcessSharedSpawnProcessTest(unittest.TestCase):
+  def setUp(self):
+    tempdir = tempfile.gettempdir()
+    for tag in ['basic',
+                'main',
+                'to_delete',
+                'mix1',
+                'mix2',
+                'test_process_exit',
+                'thundering_herd_test']:
+      for ext in ['', '.address', '.address.error']:
+        try:
+          os.remove(os.path.join(tempdir, tag + ext))
+        except OSError:
+          pass
+
+  def tearDown(self):
+    for p in multiprocessing.active_children():
+      p.terminate()
+      p.join()
+
+  def test_call(self):
+    shared = multi_process_shared.MultiProcessShared(
+        Counter, tag='basic', always_proxy=True, spawn_process=True).acquire()
+    self.assertEqual(shared.get(), 0)
+    self.assertEqual(shared.increment(), 1)
+    self.assertEqual(shared.increment(10), 11)
+    self.assertEqual(shared.increment(value=10), 21)
+    self.assertEqual(shared.get(), 21)
+
+  def test_unsafe_hard_delete_autoproxywrapper(self):
+    shared1 = multi_process_shared.MultiProcessShared(
+        Counter, tag='to_delete', always_proxy=True, spawn_process=True)
+    shared2 = multi_process_shared.MultiProcessShared(
+        Counter, tag='to_delete', always_proxy=True, spawn_process=True)
+    counter3 = multi_process_shared.MultiProcessShared(
+        Counter, tag='basic', always_proxy=True, spawn_process=True).acquire()
+
+    counter1 = shared1.acquire()
+    counter2 = shared2.acquire()
+    self.assertEqual(counter1.increment(), 1)
+    self.assertEqual(counter2.increment(), 2)
+
+    try:
+      counter2.singletonProxy_unsafe_hard_delete()
+    except Exception:
+      pass
+
+    with self.assertRaises(Exception):
+      counter1.get()
+    with self.assertRaises(Exception):
+      counter2.get()
+
+    counter4 = multi_process_shared.MultiProcessShared(
+        Counter, tag='to_delete', always_proxy=True,
+        spawn_process=True).acquire()
+
+    self.assertEqual(counter3.increment(), 1)
+    self.assertEqual(counter4.increment(), 1)
+
+  def test_mix_usage(self):
+    shared1 = multi_process_shared.MultiProcessShared(
+        Counter, tag='mix1', always_proxy=True, spawn_process=False).acquire()
+    shared2 = multi_process_shared.MultiProcessShared(
+        Counter, tag='mix2', always_proxy=True, spawn_process=True).acquire()
+
+    self.assertEqual(shared1.get(), 0)
+    self.assertEqual(shared1.increment(), 1)
+    self.assertEqual(shared2.get(), 0)
+    self.assertEqual(shared2.increment(), 1)
+
+  def test_process_exits_on_unsafe_hard_delete(self):
+    shared = multi_process_shared.MultiProcessShared(
+        Counter, tag='test_process_exit', always_proxy=True, 
spawn_process=True)
+    obj = shared.acquire()
+
+    self.assertEqual(obj.increment(), 1)
+
+    children = multiprocessing.active_children()
+    server_process = None
+    for p in children:
+      if p.pid != os.getpid() and p.is_alive():
+        server_process = p
+        break
+
+    self.assertIsNotNone(
+        server_process, "Could not find spawned server process")
+    try:
+      obj.singletonProxy_unsafe_hard_delete()
+    except Exception:
+      pass
+    server_process.join(timeout=5)
+
+    self.assertFalse(
+        server_process.is_alive(),
+        f"Server process {server_process.pid} is still alive after hard 
delete")
+    self.assertIsNotNone(
+        server_process.exitcode, "Process has no exit code (did not exit)")
+
+    with self.assertRaises(Exception):
+      obj.get()
+
+  def test_process_exits_on_unsafe_hard_delete_with_manager(self):
+    shared = multi_process_shared.MultiProcessShared(
+        Counter, tag='test_process_exit', always_proxy=True, 
spawn_process=True)
+    obj = shared.acquire()
+
+    self.assertEqual(obj.increment(), 1)
+
+    children = multiprocessing.active_children()
+    server_process = None
+    for p in children:
+      if p.pid != os.getpid() and p.is_alive():
+        server_process = p
+        break
+
+    self.assertIsNotNone(
+        server_process, "Could not find spawned server process")
+    try:
+      shared.unsafe_hard_delete()
+    except Exception:
+      pass
+    server_process.join(timeout=5)
+
+    self.assertFalse(
+        server_process.is_alive(),
+        f"Server process {server_process.pid} is still alive after hard 
delete")
+    self.assertIsNotNone(
+        server_process.exitcode, "Process has no exit code (did not exit)")
+
+    with self.assertRaises(Exception):
+      obj.get()
+
+  def test_zombie_reaping_on_acquire(self):

Review Comment:
   Hm, the test is now passing, but tox is still having issues. I'm guessing 
that we're leaving behind some process or resources which are making it 
impossible for it to close, but I'm not totally sure.
   
   Either that or we're actually sending a kill signal somewhere to the test 
process. The error is:
   
   ```
   ROOT: [2572] KeyboardInterrupt - teardown started
   ```
   
   and then we exit with error code 1



-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to