Improve logic of unexpected processes check started by agent

azurelinuxagent/common/cgroupconfigurator.py

@@ -605,7 +605,8 @@ def _check_processes_in_agent_cgroup(self):
                     current = process
                     while current != 0 and current not in agent_commands:
                         current = self._get_parent(current)
-                    if current == 0:
+                    # Process started by agent will have a marker and check if that marker found in process environment.
+                    if current == 0 and not self.__is_process_descendant_of_the_agent(process):
                         unexpected.append(self.__format_process(process))
                         if len(unexpected) >= 5:  # collect just a small sample
                             break
@@ -640,6 +641,24 @@ def __format_process(pid):
                 pass
             return "[PID: {0}] UNKNOWN".format(pid)
 
+        @staticmethod
+        def __is_process_descendant_of_the_agent(pid):
+            """
+            Returns True if the process is descendant of the agent by looking at the env flag(AZURE_GUEST_AGENT_PARENT_PROCESS_NAME)
+            that we set when the process starts otherwise False.
+            """
+            try:
+                env = '/proc/{0}/environ'.format(pid)
+                if os.path.exists(env):
+                    with open(env, "r") as env_file:
+                        environ = env_file.read()
+                        if environ and environ[-1] == '\x00':
+                            environ = environ[:-1]
+                        return "{0}={1}".format(shellutil.PARENT_PROCESS_NAME, shellutil.AZURE_GUEST_AGENT) in environ
+            except Exception:
+                pass
+            return False
+
         @staticmethod
         def _check_agent_throttled_time(cgroup_metrics):
             for metric in cgroup_metrics:

azurelinuxagent/common/utils/shellutil.py

@@ -16,7 +16,7 @@
 #
 # Requires Python 2.6+ and Openssl 1.0+
 #
-
+import os
 import subprocess
 import tempfile
 import threading
@@ -345,10 +345,23 @@ def quote(word_list):
 #
 _running_commands = []
 _running_commands_lock = threading.RLock()
+PARENT_PROCESS_NAME = "AZURE_GUEST_AGENT_PARENT_PROCESS_NAME"
+AZURE_GUEST_AGENT = "AZURE_GUEST_AGENT"
 
 
 def _popen(*args, **kwargs):
     with _running_commands_lock:
+        # Add the environment variables
+        env = {}
+        if 'env' in kwargs:
+            env.update(kwargs['env'])
+        else:
+            env.update(os.environ)
+
+        # Set the marker before process start
+        env[PARENT_PROCESS_NAME] = AZURE_GUEST_AGENT
+        kwargs['env'] = env
+
         process = subprocess.Popen(*args, **kwargs)
         _running_commands.append(process.pid)
         return process

tests/ga/test_multi_config_extension.py

@@ -738,7 +738,10 @@ def __assert_env_variables(handler_name, handler_version="1.0.0", seq_no="1", ex
                     self.assertFalse(any(env_var in commands['data'] for env_var in not_expected), "Unwanted env variable found")
 
         def mock_popen(cmd, *_, **kwargs):
-            if 'env' in kwargs:
+            # This cgroupsapi Popen mocking all other popen calls which breaking the extension emulator logic.
+            # The emulator should be used only on extension commands and not on other commands even env flag set.
+            # So, added ExtensionVersion check to avoid using extension emulator on non extension operations.
+            if 'env' in kwargs and ExtCommandEnvVariable.ExtensionVersion in kwargs['env']:
                 handler_name, __, command = extract_extension_info_from_command(cmd)
                 name = handler_name
                 if ExtCommandEnvVariable.ExtensionName in kwargs['env']:

tests/utils/test_shell_util.py

@@ -17,6 +17,7 @@
 #
 import os
 import signal
+import subprocess
 import tempfile
 import threading
 import unittest
@@ -156,6 +157,19 @@ def test_run_command_should_execute_the_command(self):
         ret = shellutil.run_command(command)
         self.assertEqual(ret, "A TEST STRING")
 
+    def test_run_command_should_use_popen_arg_list(self):
+        with patch("azurelinuxagent.common.utils.shellutil.subprocess.Popen", wraps=subprocess.Popen) as popen_patch:
+            command = ["echo", "-n", "A TEST STRING"]
+            ret = shellutil.run_command(command)
+
+            self.assertEqual(ret, "A TEST STRING")
+            self.assertEqual(popen_patch.call_count, 1)
+
+            args, kwargs = popen_patch.call_args
+            self.assertTrue(any(arg for arg in args[0] if "A TEST STRING" in arg), "command not being used")
+            self.assertEqual(kwargs['env'].get(shellutil.PARENT_PROCESS_NAME), shellutil.AZURE_GUEST_AGENT,
+                             "Env flag not being used")
+
     def test_run_pipe_should_execute_a_pipe_with_two_commands(self):
         # Output the same string 3 times and then remove duplicates
         test_string = "A TEST STRING\n"