add time window for agent manifest download

azurelinuxagent/common/conf.py

@@ -129,17 +129,17 @@ def load_conf_from_file(conf_file_path, conf=__conf__):
     "ResourceDisk.EnableSwapEncryption": False,
     "AutoUpdate.Enabled": True,
     "EnableOverProvisioning": True,
-    "GAUpdates.Enabled": True,
     #
     # "Debug" options are experimental and may be removed in later
     # versions of the Agent.
     #
     "Debug.CgroupLogMetrics": False,
     "Debug.CgroupDisableOnProcessCheckFailure": True,
     "Debug.CgroupDisableOnQuotaCheckFailure": True,
+    "Debug.DownloadNewAgents": True,
     "Debug.EnableAgentMemoryUsageCheck": False,
     "Debug.EnableFastTrack": True,
-    "Debug.EnableGAVersioning": False
+    "Debug.EnableGAVersioning": True
 }
 
 
@@ -503,12 +503,15 @@ def get_monitor_network_configuration_changes(conf=__conf__):
     return conf.get_switch("Monitor.NetworkConfigurationChanges", False)
 
 
-def get_ga_updates_enabled(conf=__conf__):
+def get_download_new_agents(conf=__conf__):
     """
-    If True, the agent go through update logic to look for new agents otherwise it will stop agent updates.
-    NOTE: This option is needed in e2e tests to control agent updates.
+    If True, the agent go through update logic to look for new agents to download otherwise it will stop agent updates.
+    NOTE: AutoUpdate.Enabled controls whether the Agent downloads new update and also whether any downloaded updates are started or not, while DownloadNewAgents controls only the former.
+    AutoUpdate.Enabled == false -> Agent preinstalled on the image will process extensions and will not update (regardless of DownloadNewAgents flag)
+    AutoUpdate.Enabled == true and DownloadNewAgents == true, any update already downloaded will be started, and agent look for future updates
+    AutoUpdate.Enabled == true and DownloadNewAgents == false, any update already downloaded will be started, but the agent will not look for future updates
     """
-    return conf.get_switch("GAUpdates.Enabled", True)
+    return conf.get_switch("Debug.DownloadNewAgents", True)
 
 
 def get_cgroup_check_period(conf=__conf__):
@@ -637,7 +640,7 @@ def get_normal_upgrade_frequency(conf=__conf__):
 
 def get_enable_ga_versioning(conf=__conf__):
     """
-    If True, the agent uses GA Versioning for auto-updating the agent vs automatically auto-updating to the highest version.
+    If True, the agent looks for rsm updates(checking requested version in GS) otherwise it will fall back to self-update and finds the highest version from PIR.
     NOTE: This option is experimental and may be removed in later versions of the Agent.
     """
     return conf.get_switch("Debug.EnableGAVersioning", True)

azurelinuxagent/ga/agent_update_handler.py

@@ -37,6 +37,7 @@ def __init__(self):
         self.last_attempted_requested_version_update_time = datetime.datetime.min
         self.last_attempted_hotfix_update_time = datetime.datetime.min
         self.last_attempted_normal_update_time = datetime.datetime.min
+        self.last_attempted_manifest_download_time = datetime.datetime.min
 
 
 class AgentUpdateHandler(object):
@@ -86,6 +87,23 @@ def __update_last_attempt_update_times(self):
         else:
             self.update_state.last_attempted_normal_update_time = now
             self.update_state.last_attempted_hotfix_update_time = now
+            self.update_state.last_attempted_manifest_download_time = now
+
+    def __should_agent_attempt_manifest_download(self):
+        """
+        The agent should attempt to download the manifest if
+        the agent has not attempted to download the manifest in the last 1 hour
+        """
+        now = datetime.datetime.now()
+
+        if self.update_state.last_attempted_manifest_download_time != datetime.datetime.min:
+            next_attempt_time = self.update_state.last_attempted_manifest_download_time + datetime.timedelta(seconds=conf.get_autoupdate_frequency())
+        else:
+            next_attempt_time = now
+
+        if next_attempt_time > now:
+            return False
+        return True
 
     @staticmethod
     def __get_agent_upgrade_type(requested_version):
@@ -256,7 +274,7 @@ def __log_event(level, msg, success=True):
     def run(self, goal_state):
         try:
             # Ignore new agents if update is disabled. The latter flag only used in e2e tests.
-            if not self._autoupdate_enabled or not conf.get_ga_updates_enabled():
+            if not self._autoupdate_enabled or not conf.get_download_new_agents():
                 return
 
             self._gs_id = goal_state.extensions_goal_state.id
@@ -265,6 +283,9 @@ def run(self, goal_state):
             agent_manifest = None  # This is to make sure fetch agent manifest once per update
             warn_msg = ""
             if requested_version is None:
+                # Do not proceed with update if self-update needs to download the manifest again with in an hour
+                if not self.__should_agent_attempt_manifest_download():
+                    return
                 if conf.get_enable_ga_versioning():  # log the warning only when ga versioning is enabled
                     warn_msg = "Missing requested version in agent family: {0} for incarnation: {1}, fallback to largest version update".format(self._ga_family, self._gs_id)
                     GAUpdateReportState.report_error_msg = warn_msg

tests/ga/test_agent_update_handler.py

@@ -28,7 +28,7 @@ def setUp(self):
         clear_singleton_instances(ProtocolUtil)
 
     @contextlib.contextmanager
-    def __get_agent_update_handler(self, test_data=None, autoupdate_frequency=0.001, autoupdate_enabled=True):
+    def __get_agent_update_handler(self, test_data=None, autoupdate_frequency=0.001, autoupdate_enabled=True, protocol_get_error=False):
         # Default to DATA_FILE of test_data parameter raises the pylint warning
         # W0102: Dangerous default value DATA_FILE (builtins.dict) as argument (dangerous-default-value)
         test_data = DATA_FILE if test_data is None else test_data
@@ -37,9 +37,12 @@ def __get_agent_update_handler(self, test_data=None, autoupdate_frequency=0.001,
 
             def get_handler(url, **kwargs):
                 if HttpRequestPredicates.is_agent_package_request(url):
-                    agent_pkg = load_bin_data(self._get_agent_file_name(), self._agent_zip_dir)
-                    protocol.mock_wire_data.call_counts['agentArtifact'] += 1
-                    return MockHttpResponse(status=httpclient.OK, body=agent_pkg)
+                    if not protocol_get_error:
+                        agent_pkg = load_bin_data(self._get_agent_file_name(), self._agent_zip_dir)
+                        return MockHttpResponse(status=httpclient.OK, body=agent_pkg)
+                    else:
+                        return MockHttpResponse(status=httpclient.SERVICE_UNAVAILABLE)
+
                 return protocol.mock_wire_data.mock_http_get(url, **kwargs)
 
             def put_handler(url, *args, **_):
@@ -59,6 +62,7 @@ def put_handler(url, *args, **_):
                             agent_update_handler._protocol = protocol
                             yield agent_update_handler, mock_telemetry
 
+
     def __assert_agent_directories_available(self, versions):
         for version in versions:
             self.assertTrue(os.path.exists(self.agent_dir(version)), "Agent directory {0} not found".format(version))
@@ -174,6 +178,33 @@ def test_it_should_update_to_largest_version_if_requested_version_not_available(
                 self.__assert_agent_directories_exist_and_others_dont_exist(versions=[str(CURRENT_VERSION), "99999.0.0.0"])
                 self.assertIn("Agent update found, exiting current process", ustr(context.exception.reason))
 
+    def test_it_should_not_download_manifest_again_if_last_attempted_download_time_not_elapsed(self):
+        self.prepare_agents(count=1)
+        data_file = DATA_FILE.copy()
+        data_file['ext_conf'] = "wire/ext_conf.xml"
+        with self.__get_agent_update_handler(test_data=data_file, autoupdate_frequency=10, protocol_get_error=True) as (agent_update_handler, _):
+            # making multiple agent update attempts
+            agent_update_handler.run(agent_update_handler._protocol.get_goal_state())
+            agent_update_handler.run(agent_update_handler._protocol.get_goal_state())
+            agent_update_handler.run(agent_update_handler._protocol.get_goal_state())
+
+            mock_wire_data = agent_update_handler._protocol.mock_wire_data
+            self.assertEqual(1, mock_wire_data.call_counts['manifest_of_ga.xml'], "Agent manifest should not be downloaded again")
+
+    def test_it_should_download_manifest_if_last_attempted_download_time_is_elapsed(self):
+        self.prepare_agents(count=1)
+        data_file = DATA_FILE.copy()
+        data_file['ext_conf'] = "wire/ext_conf.xml"
+
+        with self.__get_agent_update_handler(test_data=data_file, autoupdate_frequency=0.00001, protocol_get_error=True) as (agent_update_handler, _):
+            # making multiple agent update attempts
+            agent_update_handler.run(agent_update_handler._protocol.get_goal_state())
+            agent_update_handler.run(agent_update_handler._protocol.get_goal_state())
+            agent_update_handler.run(agent_update_handler._protocol.get_goal_state())
+
+        mock_wire_data = agent_update_handler._protocol.mock_wire_data
+        self.assertEqual(3, mock_wire_data.call_counts['manifest_of_ga.xml'], "Agent manifest should be downloaded in all attempts")
+
     def test_it_should_not_agent_update_if_requested_version_is_same_as_current_version(self):
         data_file = DATA_FILE.copy()
         data_file["ext_conf"] = "wire/ext_conf_requested_version.xml"
@@ -313,24 +344,7 @@ def test_it_should_report_update_status_with_error_on_download_fail(self):
         data_file = DATA_FILE.copy()
         data_file["ext_conf"] = "wire/ext_conf_requested_version.xml"
 
-        @contextlib.contextmanager
-        def mock_agent_update_handler(test_data):
-            with mock_wire_protocol(test_data) as protocol:
-
-                def get_handler(url, **kwargs):
-                    if HttpRequestPredicates.is_agent_package_request(url):
-                        return MockHttpResponse(status=httpclient.SERVICE_UNAVAILABLE)
-                    return protocol.mock_wire_data.mock_http_get(url, **kwargs)
-
-                protocol.set_http_handlers(http_get_handler=get_handler)
-
-                with patch("azurelinuxagent.common.conf.get_autoupdate_enabled", return_value=True):
-                    with patch("azurelinuxagent.common.conf.get_autoupdate_frequency", return_value=0.001):
-                        with patch("azurelinuxagent.common.conf.get_autoupdate_gafamily", return_value="Prod"):
-                            agent_update_handler_local = get_agent_update_handler(protocol)
-                            yield agent_update_handler_local
-
-        with mock_agent_update_handler(test_data=data_file) as (agent_update_handler):
+        with self.__get_agent_update_handler(test_data=data_file, protocol_get_error=True) as (agent_update_handler, _):
             GAUpdateReportState.report_error_msg = ""
             agent_update_handler.run(agent_update_handler._protocol.get_goal_state())
             vm_agent_update_status = agent_update_handler.get_vmagent_update_status()
@@ -343,23 +357,7 @@ def test_it_should_report_update_status_with_missing_requested_version_error(sel
         data_file = DATA_FILE.copy()
         data_file['ext_conf'] = "wire/ext_conf.xml"
 
-        @contextlib.contextmanager
-        def mock_agent_update_handler(test_data):
-            with mock_wire_protocol(test_data) as protocol:
-                def get_handler(url, **kwargs):
-                    if HttpRequestPredicates.is_agent_package_request(url):
-                        return MockHttpResponse(status=httpclient.SERVICE_UNAVAILABLE)
-                    return protocol.mock_wire_data.mock_http_get(url, **kwargs)
-
-                protocol.set_http_handlers(http_get_handler=get_handler)
-
-                with patch("azurelinuxagent.common.conf.get_autoupdate_enabled", return_value=True):
-                    with patch("azurelinuxagent.common.conf.get_autoupdate_frequency", return_value=0.001):
-                        with patch("azurelinuxagent.common.conf.get_autoupdate_gafamily", return_value="Prod"):
-                            agent_update_handler_local = get_agent_update_handler(protocol)
-                            yield agent_update_handler_local
-
-        with mock_agent_update_handler(test_data=data_file) as (agent_update_handler):
+        with self.__get_agent_update_handler(test_data=data_file, protocol_get_error=True) as (agent_update_handler, _):
             GAUpdateReportState.report_error_msg = ""
             agent_update_handler.run(agent_update_handler._protocol.get_goal_state())
             vm_agent_update_status = agent_update_handler.get_vmagent_update_status()

tests/test_agent.py

@@ -40,9 +40,10 @@
 Debug.CgroupLogMetrics = False
 Debug.CgroupMonitorExpiryTime = 2022-03-31
 Debug.CgroupMonitorExtensionName = Microsoft.Azure.Monitor.AzureMonitorLinuxAgent
+Debug.DownloadNewAgents = True
 Debug.EnableAgentMemoryUsageCheck = False
 Debug.EnableFastTrack = True
-Debug.EnableGAVersioning = False
+Debug.EnableGAVersioning = True
 Debug.EtpCollectionPeriod = 300
 Debug.FirewallRulesLogPeriod = 86400
 DetectScvmmEnv = False
@@ -51,7 +52,6 @@
 Extensions.Enabled = True
 Extensions.GoalStatePeriod = 6
 Extensions.InitialGoalStatePeriod = 6
-GAUpdates.Enabled = True
 HttpProxy.Host = None
 HttpProxy.Port = None
 Lib.Dir = /var/lib/waagent

tests_e2e/orchestrator/scripts/install-agent

@@ -115,7 +115,7 @@ echo "Agent's conf path: $waagent_conf_path"
 sed -i 's/AutoUpdate.Enabled=n/AutoUpdate.Enabled=y/g' "$waagent_conf_path"
 # By default GAUpdates flag set to True, so that agent go through update logic to look for new agents.
 # But in e2e tests this flag needs to be off in test version 9.9.9.9 to stop the agent updates, so that our scenarios run on 9.9.9.9.
-sed -i '$a GAUpdates.Enabled=n' "$waagent_conf_path"
+sed -i '$a Debug.DownloadNewAgents=n' "$waagent_conf_path"
 
 #
 # Restart the service