Fixed dead lock in telemetry (#4873)

* Fixed dead lock in telemetry

* Refactored TelemetrySender.send function

* Refactored send function implementation to avoid deadlocks

* Unit tests for telemetry sender function

* Added legal header

model-optimizer/telemetry/utils/sender.py

@@ -34,13 +34,18 @@ class TelemetrySender:
             with self._lock:
                 self.queue_size -= 1
 
+        free_space = False
         with self._lock:
             if self.queue_size < MAX_QUEUE_SIZE:
-                fut = self.executor.submit(backend.send, message)
-                fut.add_done_callback(_future_callback)
+                free_space = True
                 self.queue_size += 1
             else:
                 pass  # dropping a message because the queue is full
+        # to avoid dead lock we should not add callback inside the "with self._lock" block because it will be executed
+        # immediately if the fut is available
+        if free_space:
+            fut = self.executor.submit(backend.send, message)
+            fut.add_done_callback(_future_callback)
 
     def force_shutdown(self, timeout: float):
         """
@@ -53,11 +58,14 @@ class TelemetrySender:
         :return: None
         """
         try:
+            need_sleep = False
             with self._lock:
                 if self.queue_size > 0:
-                    sleep(timeout)
-                self.executor.shutdown(wait=False)
-                self.executor._threads.clear()
-                futures.thread._threads_queues.clear()
+                    need_sleep = True
+            if need_sleep:
+                sleep(timeout)
+            self.executor.shutdown(wait=False)
+            self.executor._threads.clear()
+            futures.thread._threads_queues.clear()
         except Exception:
             pass

model-optimizer/telemetry/utils/sender_test.py

@@ -0,0 +1,81 @@
+"""
+ Copyright (C) 2018-2021 Intel Corporation
+
+ Licensed under the Apache License, Version 2.0 (the "License");
+ you may not use this file except in compliance with the License.
+ You may obtain a copy of the License at
+
+      http://www.apache.org/licenses/LICENSE-2.0
+
+ Unless required by applicable law or agreed to in writing, software
+ distributed under the License is distributed on an "AS IS" BASIS,
+ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ See the License for the specific language governing permissions and
+ limitations under the License.
+"""
+
+import unittest
+import time
+
+from telemetry.utils.sender import TelemetrySender
+
+
+class FakeTelemetryBackend:
+    def send(self, param):
+        pass
+
+
+class FakeTelemetryBackendWithSleep:
+    def send(self, param):
+        time.sleep(1)
+
+
+class TelemetrySenderStress(unittest.TestCase):
+    def test_stress(self):
+        """
+        Stress tests to schedule a lot of threads which works super fast (do nothing) with sending messages."
+        """
+        tm = TelemetrySender()
+        fake_backend = FakeTelemetryBackend()
+        for _ in range(1000000):
+            tm.send(fake_backend, None)
+
+    def test_check_shutdown(self):
+        """
+        Stress test to schedule many threads taking 1 second and then ask to force shutdown. Make sure that the elapsed
+        time is small.
+        """
+        tm = TelemetrySender()
+        fake_backend = FakeTelemetryBackendWithSleep()
+        # schedule many requests which just wait 1 second
+        for _ in range(100000):
+            tm.send(fake_backend, None)
+
+        start_time = time.time()
+        # ask to shutdown with timeout of 1 second
+        tm.force_shutdown(1)
+        while len(tm.executor._threads):
+            pass
+        # check that no more than 3 seconds spent
+        self.assertTrue(time.time() - start_time < 3)
+
+    def test_check_shutdown_negative(self):
+        """
+        Test to check that without forcing shutdown total execution time is expected.
+        """
+        tm = TelemetrySender(1)  # only one worker thread
+        fake_backend = FakeTelemetryBackendWithSleep()
+        start_time = time.time()
+        # schedule 5 requests which totally should work more than 4 seconds
+        for _ in range(5):
+            tm.send(fake_backend, None)
+
+        try:
+            # wait until all threads finish their work. We use internal ThreadPoolExecutor attribute _work_queue to make
+            # sure that all workers completed their work, so the whole code is wrapped to try/except to avoid exceptions
+            # if internal implementation is changed in the future
+            while tm.executor._work_queue.qsize():
+                pass
+            self.assertTrue(time.time() - start_time > 4.0)
+        except:
+            pass