[AMA] Improve the uninstall of AMA by removing files for clean uninstall situation (#2087)

* cleanup of files for a clean uninstall situation. uninstall command is unaffected

* wrong commit

* refactoring to support current operations and have the update operation cache files for install later on

* cleanup

* preserving log files into cache. may need to look into rpm vs dpkg due to difference in uninstall command

* Removed events folder for extension logs. Removes /opt/microsoft/azuremonitoragent for rpm. Caches logs in /var/opt/microsoft/azuremonitoragent/log

* cleanup

* fixed path

* added caching of tenants

* more robust file and directory collection for uninstall

* Config files such as tenants are cached, removing changes related to this.

* Changed patternt o be use context from a marker file

* add comments and update docstring

* fixed logging for debug and edited comments

* updated doc on update()

* quick change to bundlefilename

* added logging and tested for python 2 and 3 compat

* spacing

* code cleanup and changes based on feedback

* minor cleanup and moved call to get_uninstall_context()

* spacing

* updated comments and logging

* remove extra period

* cleanup code, style changes

* cleanup

* added constant for AMA data path

---------

Co-authored-by: Dylan Bunarto <[email protected]>

Author: dylanbun

AzureMonitorAgent/agent.py

@@ -42,9 +42,7 @@
 import ama_tst.modules.install.supported_distros as supported_distros
 from collections import OrderedDict
 from hashlib import sha256
-from shutil import copyfile
-from shutil import copytree
-from shutil import rmtree
+from shutil import copyfile, rmtree, copytree, copy2
 
 from threading import Thread
 import telegraf_utils.telegraf_config_handler as telhandler
@@ -134,6 +132,8 @@ def __str__(self):
 AMAAstTransformConfigMarkerPath = '/etc/opt/microsoft/azuremonitoragent/config-cache/agenttransform.marker'
 AMAExtensionLogRotateFilePath = '/etc/logrotate.d/azuremonitoragentextension'
 WAGuestAgentLogRotateFilePath = '/etc/logrotate.d/waagent-extn.logrotate'
+AmaUninstallContextFile = '/var/opt/microsoft/uninstall-context'
+AmaDataPath = '/var/opt/microsoft/azuremonitoragent/'
 SupportedArch = set(['x86_64', 'aarch64'])
 
 # Error codes
@@ -538,7 +538,7 @@ def install():
 def uninstall():
     """
     Uninstall the Azure Monitor Linux Agent.
-    This is a somewhat soft uninstall. It is not a purge.
+    Whether it is a purge of all files or preserve of log files depends on the uninstall context file.
     Note: uninstall operation times out from WAAgent at 5 minutes
     """
 
@@ -554,6 +554,13 @@ def uninstall():
     if PackageManager != "dpkg" and PackageManager != "rpm":
         log_and_exit("Uninstall", UnsupportedOperatingSystem, "The OS has neither rpm nor dpkg." )
 
+    # For clean uninstall, gather the file list BEFORE running the uninstall command
+    # This ensures we have the complete list even after the package manager removes its database
+    package_files_for_cleanup = []
+
+    hutil_log_info("Gathering package file list for clean uninstall before removing package")
+    package_files_for_cleanup = _get_package_files_for_cleanup()
+        
     # Attempt to uninstall each specific
 
     # Try a specific package uninstall for rpm
@@ -604,6 +611,13 @@ def uninstall():
     # Retry, since uninstall can fail due to concurrent package operations
     try:
         exit_code, output = force_uninstall_azure_monitor_agent()
+
+        # Remove all files installed by the package that were listed
+        _remove_package_files_from_list(package_files_for_cleanup)
+
+        # Clean up context marker (always do this)
+        _cleanup_uninstall_context()
+
     except Exception as ex:
         exit_code = GenericErrorCode
         output = 'Uninstall failed with error: {0}\n' \
@@ -683,6 +697,110 @@ def force_uninstall_azure_monitor_agent():
     else:
         hutil_log_info("Azure Monitor Agent has been uninstalled.")
         return 0, "Azure Monitor Agent has been uninstalled."
+      
+def _get_package_files_for_cleanup():
+    """
+    Get the list of files and directories installed by the provided
+    azuremonitoragent spec that should be removed during uninstall.
+    This must be called BEFORE the package is uninstalled to ensure the package
+    manager still has the file list available.
+    
+    Returns:
+        tuple: (files_list, directories_to_add) where files_list contains package files
+               and directories_to_add contains directories that need explicit cleanup
+    """
+    try:
+        # Get list of files installed by the package
+        if PackageManager == "dpkg":
+            # For Debian-based systems
+            cmd = "dpkg -L azuremonitoragent"
+        elif PackageManager == "rpm":
+            # For RPM-based systems
+            cmd = "rpm -ql azuremonitoragent"
+        else:
+            hutil_log_info("Unknown package manager, cannot list package files")
+            return []
+
+        exit_code, output = run_command_and_log(cmd, check_error=False)
+        
+        if exit_code != 0 or not output:
+            hutil_log_info("Could not get package file list for cleanup")
+            return []
+
+        # Parse the file list
+        files = [line.strip() for line in output.strip().split('\n') if line.strip()]
+        
+        # Collect all azuremonitor-related paths
+        azuremonitoragent_files = []
+        
+        for file_path in files:
+            # Only include files/directories that have "azuremonitor" in their path
+            # This covers both "azuremonitoragent" and "azuremonitor-*" service files
+            if "azuremonitor" in file_path:
+                azuremonitoragent_files.append(file_path)
+            else:
+                hutil_log_info("Skipping non-azuremonitor path: {0}".format(file_path))
+        
+        return azuremonitoragent_files
+        
+    except Exception as ex:
+        hutil_log_error("Error gathering package files for cleanup: {0}\n Is Azure Monitor Agent Installed?".format(ex))
+        return []
+
+def _remove_package_files_from_list(package_files):
+    """
+    Remove all files and directories from the provided list that were installed 
+    by the provided azuremonitoragent spec. This function works with a pre-gathered 
+    list of files from _get_package_files_for_cleanup(), allowing it to work even 
+    after the package has been uninstalled.
+    
+    Args:
+        package_files (list): List of file/directory paths to remove
+    """
+    try:
+        if not package_files:
+            hutil_log_info("No package files provided for removal")
+            return
+            
+        # Build consolidated list of paths to clean up
+        cleanup_paths = set(package_files) if package_files else set()
+        
+        # Add directories that need explicit cleanup since on rpm systems 
+        # the initial list for this path does not remove the directories and files
+        cleanup_paths.add("/opt/microsoft/azuremonitoragent/")
+
+        # Determine uninstall context based on if the context file exists
+        uninstall_context = _get_uninstall_context()
+        hutil_log_info("Uninstall context: {0}".format(uninstall_context))
+        
+        if uninstall_context == 'complete':
+            hutil_log_info("Complete uninstall context - removing everything")
+            cleanup_paths.add(AmaDataPath)
+
+        # Sort paths by depth (deepest first) to avoid removing parent before children
+        sorted_paths = sorted(cleanup_paths, key=lambda x: x.count('/'), reverse=True)
+        
+        hutil_log_info("Removing {0} azuremonitor paths".format(len(sorted_paths)))
+        
+        items_removed = 0
+        for item_path in sorted_paths:
+            try:
+                if os.path.exists(item_path):
+                    if os.path.isdir(item_path):
+                        rmtree(item_path)
+                        hutil_log_info("Removed directory: {0}".format(item_path))
+                    else:
+                        os.remove(item_path)
+                        hutil_log_info("Removed file: {0}".format(item_path))
+                    items_removed += 1
+            except Exception as ex:
+                hutil_log_info("Failed to remove {0}: {1}".format(item_path, ex))
+        
+        hutil_log_info("Removed {0} items total".format(items_removed))
+        
+    except Exception as ex:
+        hutil_log_error("Error during file removal from list: {0}\n Were these files removed already?".format(ex))
+
 def enable():
     """
     Start the Azure Monitor Linux Agent Service
@@ -1116,12 +1234,63 @@ def disable():
 
 def update():
     """
-    Update the current installation of AzureMonitorLinuxAgent
-    No logic to install the agent as agent -> install() will be called
-    with update because upgradeMode = "UpgradeWithInstall" set in HandlerManifest
+    This function is called when the extension is updated.
+    It marks the uninstall context to indicate that the next run should be treated as an update rather than a clean install.
+
+    Always returns 0
     """
 
-    return 0, ""
+    hutil_log_info("Update operation called for Azure Monitor Agent")
+
+    try:
+        state_dir = os.path.dirname(AmaUninstallContextFile)
+        if not os.path.exists(state_dir):
+            os.makedirs(state_dir)
+        with open(AmaUninstallContextFile, 'w') as f:
+            f.write('update\n')
+            f.write(datetime.datetime.now().strftime('%Y-%m-%d %H:%M:%S')) # Timestamp for debugging
+        hutil_log_info("Marked uninstall context as 'update'")
+    except Exception as ex:
+        hutil_log_error("Failed to set uninstall context: {0}\n The uninstall operation will not behave as expected with the uninstall context file missing, defaulting to an uninstall that removes {1}.".format(ex, AmaDataPath))
+
+    return 0, "Update succeeded"
+
+def _get_uninstall_context():
+    """
+    Determine the context of this uninstall operation
+
+    Returns the context as a string:
+        'complete' - if this is a clean uninstall
+        'update' - if this is an update operation
+    Also returns as 'complete' if it fails to read the context file.
+    """
+
+    try:
+        if os.path.exists(AmaUninstallContextFile):
+            with open(AmaUninstallContextFile, 'r') as f:
+                context = f.read().strip().split('\n')[0]
+                hutil_log_info("Found uninstall context: {0}".format(context))
+                return context
+        else:
+            hutil_log_info("Uninstall context file does not exist, defaulting to 'complete'")
+    except Exception as ex:
+        hutil_log_error("Failed to read uninstall context file: {0}\n The uninstall operation will not behave as expected with the uninstall context file missing, defaulting to an uninstall that removes {1}.".format(ex, AmaDataPath))
+
+    return 'complete'
+
+def _cleanup_uninstall_context():
+    """
+    Clean up uninstall context marker
+    """
+
+    try:
+        if os.path.exists(AmaUninstallContextFile):
+            os.remove(AmaUninstallContextFile)
+            hutil_log_info("Removed uninstall context file")
+        else:
+            hutil_log_info("Uninstall context file does not exist, nothing to remove")
+    except Exception as ex:
+        hutil_log_error("Failed to cleanup uninstall context: {0}\n This may result in unintended behavior as described.\nIf the marker file exists and cannot be removed, uninstall will continue to keep the {1} path, leading users to have to remove it manually.".format(ex, AmaDataPath))
 
 def restart_launcher():
     # start agent launcher