Update unhealthy unload message

Author: kthui

ci/L0_check_health_vllm/check_health_test.py

@@ -108,7 +108,7 @@ def test_vllm_not_healthy(self):
         self._assert_model_ready(True)
         # The 2nd infer should begin with health check failed
         self._llm_infer()
-        self._assert_infer_exception("vLLM engine is not healthy")
+        self._assert_infer_exception("vLLM engine is not healthy and will be unloaded")
         self._assert_model_ready(False)
         # The 3rd infer should have model not found
         self._llm_infer()

src/model.py

@@ -558,7 +558,9 @@ def _check_health(self, requests):
         try:
             future.result()
         except Exception as e:
-            self.logger.log_error(f"[vllm] Engine is not healthy: {e}")
+            self.logger.log_error(
+                f"[vllm] Engine is not healthy and will be unloaded: {e}"
+            )
             pb_utils.unload_model(self.model_config["name"])  # non-blocking
             self._is_healthy = False
         if not self._is_healthy: