Store process UID in the database; use it to determine the parent/chi…

…ld process. Stop relying on PIDs when detecting running processes

src/run_tribler_headless.py

@@ -81,6 +81,7 @@ async def signal_handler(sig):
         current_process = TriblerProcess.current_process(ProcessKind.Core)
         self.process_manager = ProcessManager(root_state_dir, current_process)
         set_global_process_manager(self.process_manager)
+        current_process.start_updating_thread()
 
         if not self.process_manager.current_process.become_primary():
             msg = 'Another Core process is already running'

src/tribler/core/components/gui_process_watcher/gui_process_watcher.py

@@ -5,6 +5,7 @@
 import psutil
 from ipv8.taskmanager import TaskManager
 
+GUI_UID_ENV_KEY = 'TRIBLER_GUI_UID'
 GUI_PID_ENV_KEY = 'TRIBLER_GUI_PID'
 CHECK_INTERVAL = 10
 
@@ -52,6 +53,16 @@ def get_gui_pid() -> Optional[int]:
                 logger.warning(f'Cannot parse {GUI_PID_ENV_KEY} environment variable: {pid}')
         return None
 
+    @staticmethod
+    def get_gui_uid() -> Optional[int]:
+        uid = os.environ.get(GUI_UID_ENV_KEY, None)
+        if uid:
+            try:
+                return int(uid)
+            except ValueError:
+                logger.warning(f'Cannot parse {GUI_UID_ENV_KEY} environment variable: {uid}')
+        return None
+
     @classmethod
     def get_gui_process(cls) -> Optional[psutil.Process]:
         pid = cls.get_gui_pid()

src/tribler/core/components/gui_process_watcher/tests/test_gui_process_watcher.py

@@ -4,7 +4,8 @@
 import psutil
 import pytest
 
-from tribler.core.components.gui_process_watcher.gui_process_watcher import GUI_PID_ENV_KEY, GuiProcessNotRunning, \
+from tribler.core.components.gui_process_watcher.gui_process_watcher import GUI_PID_ENV_KEY, GUI_UID_ENV_KEY, \
+    GuiProcessNotRunning, \
     GuiProcessWatcher
 
 
@@ -33,6 +34,19 @@ def test_get_gui_pid(caplog):
         assert GuiProcessWatcher.get_gui_pid() == 123
 
 
+def test_get_gui_uid(caplog):
+    with patch.dict(os.environ, {GUI_UID_ENV_KEY: ''}):
+        assert GuiProcessWatcher.get_gui_pid() is None
+
+    with patch.dict(os.environ, {GUI_UID_ENV_KEY: 'abc'}):
+        caplog.clear()
+        assert GuiProcessWatcher.get_gui_uid() is None
+        assert caplog.records[-1].message == 'Cannot parse TRIBLER_GUI_UID environment variable: abc'
+
+    with patch.dict(os.environ, {GUI_UID_ENV_KEY: '123'}):
+        assert GuiProcessWatcher.get_gui_uid() == 123
+
+
 def test_get_gui_process():
     # pid is not specified
     with patch.dict(os.environ, {GUI_PID_ENV_KEY: ''}):

src/tribler/core/start_core.py

@@ -186,8 +186,9 @@ def run_tribler_core_session(api_port: Optional[int], api_key: str,
 def run_core(api_port: Optional[int], api_key: Optional[str], root_state_dir, parsed_args):
     logger.info(f"Running Core in {'gui_test_mode' if parsed_args.gui_test_mode else 'normal mode'}")
 
+    gui_uid = GuiProcessWatcher.get_gui_uid()
     gui_pid = GuiProcessWatcher.get_gui_pid()
-    current_process = TriblerProcess.current_process(ProcessKind.Core, creator_pid=gui_pid)
+    current_process = TriblerProcess.current_process(kind=ProcessKind.Core, creator_uid=gui_uid, creator_pid=gui_pid)
     process_manager = ProcessManager(root_state_dir, current_process)
     set_global_process_manager(process_manager)
     current_process_is_primary = process_manager.current_process.become_primary()
@@ -199,6 +200,7 @@ def run_core(api_port: Optional[int], api_key: Optional[str], root_state_dir, pa
         logger.warning(msg)
         process_manager.sys_exit(1, msg)
 
+    current_process.start_updating_thread()
     version_history = VersionHistory(root_state_dir)
     state_dir = version_history.code_version.directory
     exit_code = run_tribler_core_session(api_port, api_key, state_dir, gui_test_mode=parsed_args.gui_test_mode)

src/tribler/core/utilities/process_manager/manager.py

@@ -105,7 +105,7 @@ def _unable_to_open_db_file_get_reason(self):
 
         return 'unknown reason'
 
-    def primary_process_rowid(self, kind: ProcessKind) -> Optional[int]:
+    def primary_process_uid(self, kind: ProcessKind) -> Optional[int]:
         """
         A helper method to load the current primary process of the specified kind from the database.
 
@@ -114,18 +114,26 @@ def primary_process_rowid(self, kind: ProcessKind) -> Optional[int]:
         with self.connect() as connection:
             cursor = connection.execute(f"""
                 SELECT {sql_scripts.SELECT_COLUMNS}
-                FROM processes WHERE kind = ? and "primary" = 1 ORDER BY rowid DESC LIMIT 1
+                FROM processes WHERE kind = ? and is_primary = 1 ORDER BY rowid
             """, [kind.value])
-            row = cursor.fetchone()
-            if row is not None:
+            rows = cursor.fetchall()
+            primary_processes = []
+            for row in rows:
                 process = TriblerProcess.from_row(self, row)
                 if process.is_running():
-                    return process.rowid
+                    primary_processes.append(process)
+                else:
+                    # Process is not running anymore; mark it as not primary
+                    process.is_primary = False
+                    process.save()
 
-                # Process is not running anymore; mark it as not primary
-                process.primary = False
-                process.save()
-            return None
+            if not primary_processes:
+                return None
+
+            if len(primary_processes) > 1:
+                raise RuntimeError(f'Several primary processes found of kind {kind}')
+
+            return primary_processes[0].uid
 
     def sys_exit(self, exit_code: Optional[int] = None, error: Optional[str | Exception] = None, replace: bool = False):
         """
@@ -141,7 +149,7 @@ def sys_exit(self, exit_code: Optional[int] = None, error: Optional[str | Except
         sys.exit(exit_code)
 
     @with_retry
-    def get_last_processes(self, limit=6) -> List[TriblerProcess]:
+    def get_last_processes(self, limit=10) -> List[TriblerProcess]:
         """
         Returns last `limit` processes from the database. They are used during the formatting of the error report.
         """