Fullstack: Wait placement process fixtrue to really stop

It seems that in test test_configurations_are_synced_towards_placement
when the fake placement process is stopped it can take longer and
the test restarts the agent before it is really dead.
In this case neutron-server can report the status and the agent's
'resources_synced' field is True when the tests checks it.

A simple workaround for this can be to check if the AsyncProcess
class' _is_running property is False.

Closes-Bug: #1856319
Change-Id: Ia4c25110ee5a7b7539407d3531f559555fec50cc

neutron/agent/common/async_process.py

@@ -95,6 +95,10 @@ class AsyncProcess(object):
     def cmd(self):
         return ' '.join(self._cmd)
 
+    @property
+    def is_running(self):
+        return self._is_running
+
     def _reset_queues(self):
         self._stdout_lines = eventlet.queue.LightQueue()
         self._stderr_lines = eventlet.queue.LightQueue()

neutron/tests/fullstack/resources/process.py

@@ -98,6 +98,12 @@ class ProcessFixture(fixtures.Fixture):
         else:
             return executor.submit(_restart)
 
+    def process_is_running(self):
+        return self.process.is_running
+
+    def process_is_not_running(self):
+        return not self.process_is_running()
+
 
 class RabbitmqEnvironmentFixture(fixtures.Fixture):
 

neutron/tests/fullstack/test_agent_bandwidth_report.py

@@ -237,6 +237,14 @@ class TestPlacementBandwidthReport(base.BaseFullStackTestCase):
             sleep=1)
 
         self.environment.placement.process_fixture.stop()
+
+        placement_fixture = self.environment.placement.process_fixture
+        utils.wait_until_true(
+            predicate=functools.partial(
+                placement_fixture.process_is_not_running),
+            timeout=report_interval, sleep=1
+        )
+
         _add_new_bridge_and_restart_agent(self.environment.hosts[0])
 
         check_agent_not_synced = functools.partial(