Merge branch 'develop' into pick-disbale-GAupdate

azurelinuxagent/ga/update.py

@@ -145,7 +145,8 @@ def __init__(self):
         self._heartbeat_id = str(uuid.uuid4()).upper()
         self._heartbeat_counter = 0
 
-        self._last_check_memory_usage = datetime.min
+        self._initial_attempt_check_memory_usage = True
+        self._last_check_memory_usage_time = time.time()
         self._check_memory_usage_last_error_report = datetime.min
 
         # VM Size is reported via the heartbeat, default it here.
@@ -1016,8 +1017,11 @@ def _check_agent_memory_usage(self):
         """
         try:
             if conf.get_enable_agent_memory_usage_check() and self._extensions_summary.converged:
-                if self._last_check_memory_usage == datetime.min or datetime.utcnow() >= (self._last_check_memory_usage + UpdateHandler.CHECK_MEMORY_USAGE_PERIOD):
-                    self._last_check_memory_usage = datetime.utcnow()
+                # we delay first attempt memory usage check, so that current agent won't get blacklisted due to multiple restarts(because of memory limit reach) too frequently
+                if (self._initial_attempt_check_memory_usage and time.time() - self._last_check_memory_usage_time > CHILD_LAUNCH_INTERVAL) or \
+                        (not self._initial_attempt_check_memory_usage and time.time() - self._last_check_memory_usage_time > conf.get_cgroup_check_period()):
+                    self._last_check_memory_usage_time = time.time()
+                    self._initial_attempt_check_memory_usage = False
                     CGroupConfigurator.get_instance().check_agent_memory_usage()
         except AgentMemoryExceededException as exception:
             msg = "Check on agent memory usage:\n{0}".format(ustr(exception))

tests/ga/test_update.py

@@ -2380,7 +2380,7 @@ def test_check_agent_memory_usage_raises_exit_exception(self, patch_add_event, p
             with patch('azurelinuxagent.common.conf.get_enable_agent_memory_usage_check', return_value=True):
                 with self.assertRaises(ExitException) as context_manager:
                     update_handler = get_update_handler()
-
+                    update_handler._last_check_memory_usage_time = time.time() - 24 * 60
                     update_handler._check_agent_memory_usage()
                     self.assertEqual(1, patch_add_event.call_count)
                     self.assertTrue(any("Check on agent memory usage" in call_args[0]
@@ -2395,7 +2395,7 @@ def test_check_agent_memory_usage_fails(self, patch_add_event, patch_warn, *_):
         with patch("azurelinuxagent.ga.cgroupconfigurator.CGroupConfigurator._Impl.check_agent_memory_usage", side_effect=Exception()):
             with patch('azurelinuxagent.common.conf.get_enable_agent_memory_usage_check', return_value=True):
                 update_handler = get_update_handler()
-
+                update_handler._last_check_memory_usage_time = time.time() - 24 * 60
                 update_handler._check_agent_memory_usage()
                 self.assertTrue(any("Error checking the agent's memory usage" in call_args[0]
                                     for call_args in patch_warn.call_args),
@@ -2411,6 +2411,15 @@ def test_check_agent_memory_usage_fails(self, patch_add_event, patch_warn, *_):
                     add_events[0]["message"],
                     "The error message is not correct when memory usage check failed")
 
+    @patch("azurelinuxagent.ga.cgroupconfigurator.CGroupConfigurator._Impl.check_agent_memory_usage")
+    @patch("azurelinuxagent.ga.update.add_event")
+    def test_check_agent_memory_usage_not_called(self, patch_add_event, patch_memory_usage, *_):
+        # This test ensures that agent not called immediately on startup, instead waits for CHILD_LAUNCH_INTERVAL
+        with patch('azurelinuxagent.common.conf.get_enable_agent_memory_usage_check', return_value=True):
+            update_handler = get_update_handler()
+            update_handler._check_agent_memory_usage()
+            self.assertEqual(0, patch_memory_usage.call_count)
+            self.assertEqual(0, patch_add_event.call_count)
 
 class GoalStateIntervalTestCase(AgentTestCase):
     def test_initial_goal_state_period_should_default_to_goal_state_period(self):