fix

AlibabaPAI · Oct 9, 2024 · 99726ae · 99726ae
1 parent d7ca3a6
commit 99726ae
Show file tree

Hide file tree

Showing 2 changed files with 8 additions and 0 deletions.
diff --git a/llumnix/backends/vllm/llm_engine.py b/llumnix/backends/vllm/llm_engine.py
@@ -241,6 +241,7 @@ def __init__(
 
         self.state_lock = threading.Lock()
         self.state = EngineState.INIT
+        logger.info("{} current state {}".format(self.instance_id, self.state))
 
         self._stop_event = threading.Event()
         self._thread = threading.Thread(
@@ -252,7 +253,9 @@ def _start_engine_loop(self) -> None:
         self._stop_event.clear()
 
         with self.state_lock:
+            previous_state = self.state
             self.state = EngineState.RUNNING
+            logger.info("{} change state: {} -> {}".format(self.instance_id, previous_state, self.state))
 
         while not self._stop_event.is_set():
             try:
@@ -262,13 +265,17 @@ def _start_engine_loop(self) -> None:
                 logger.error("Error in engine loop: {}".format(e))
                 logger.error("exception traceback: {}".format(traceback.format_exc()))
                 self._run_workers("shutdown")
+
                 with self.state_lock:
+                    previous_state = self.state
                     self.state = EngineState.CRASHED
+                    logger.info("{} change state: {} -> {}".format(self.instance_id, previous_state, self.state))
                 break
 
         with self.state_lock:
             if self.state == EngineState.RUNNING:
                 self.state = EngineState.STOPPED
+                logger.info("{} change state: {} -> {}".format(self.instance_id, EngineState.RUNNING, self.state))
 
     def execute_worker_method(self, method, *args, **kwargs):
         return self.engine.model_executor.driver_worker.execute_method(method, *args, **kwargs)

diff --git a/llumnix/llumlet/llumlet.py b/llumnix/llumlet/llumlet.py
@@ -109,6 +109,7 @@ def check_state(self):
 
             with self.backend_engine.state_lock:
                 if self.backend_engine.state == EngineState.CRASHED:
+                    logger.warning("llumlet({}) detected backend engine crashed. Stopping...".format(self.instance_id))
                     # pylint: disable=protected-access
                     self.backend_engine._stop_event.set()
                     if self.backend_engine._thread.is_alive():