[TRTLLM-8507][fix] Fix ray resource cleanup and error handling in LoRA test (#8175)

Signed-off-by: shuyix <219646547+shuyixiong@users.noreply.github.com>

tensorrt_llm/_ray_utils.py

@@ -0,0 +1,28 @@
+# SPDX-FileCopyrightText: Copyright (c) 2022-2024 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+from contextlib import contextmanager
+
+try:
+    import ray
+except ImportError:
+    import tensorrt_llm.ray_stub as ray
+
+
+@contextmanager
+def unwrap_ray_errors():
+    try:
+        yield
+    except ray.exceptions.RayTaskError as e:
+        raise e.as_instanceof_cause() from e

tensorrt_llm/executor/ray_executor.py

@@ -12,6 +12,7 @@ from ray.util.placement_group import (PlacementGroup,
                                       get_current_placement_group,
                                       placement_group)
 
+from tensorrt_llm._ray_utils import unwrap_ray_errors
 from tensorrt_llm._utils import get_free_port
 from tensorrt_llm.logger import logger
 
@@ -57,6 +58,7 @@ class RayExecutor(GenerationExecutor):
             "runtime_env": runtime_env
         }
 
+        try:
             if os.environ.get("TLLM_RAY_FORCE_LOCAL_CLUSTER", "0") != "1":
                 try:
                     ray.init(address="auto", **ray_init_args)
@@ -99,6 +101,11 @@ class RayExecutor(GenerationExecutor):
                                  kv_connector_config=kv_connector_config)
 
             self.create_workers(RayGPUWorker, worker_kwargs)
+        except Exception as e:
+            # Clean up the Ray resources early during exception
+            self.shutdown()
+            logger.error(f"Failed to initialize RayExecutor: {e}")
+            raise e
 
     @staticmethod
     def create_actor_weak_ref(actor_handle: ray.actor.ActorHandle):
@@ -137,12 +144,19 @@ class RayExecutor(GenerationExecutor):
             for rank in range(self.world_size)
         ]
 
+        try:
             ray.get([worker.__ray_ready__.remote() for worker in self.workers])
+        except ray.exceptions.ActorDiedError as e:
+            if "The actor died because of an error raised in its creation task" in str(
+                    e):
+                raise RuntimeError(
+                    "RayGPUWorker died during initialization") from e
+            raise
 
+    @unwrap_ray_errors()
     def call_all_ray_workers(self, func: str, leader_only: bool,
                              async_call: bool, *args, **kwargs):
         workers = (self.workers[0], ) if leader_only else self.workers
-
         if async_call:
             return [
                 getattr(worker, func).remote(*args, **kwargs)
@@ -154,6 +168,7 @@ class RayExecutor(GenerationExecutor):
                 for worker in workers
             ])
 
+    @unwrap_ray_errors()
     def collective_rpc(self,
                        method: str,
                        args: tuple = (),
@@ -174,7 +189,6 @@ class RayExecutor(GenerationExecutor):
                 # Ray actor doesn't work with __getattr__ delegation.
                 refs.append(w.call_worker_method.remote(method, *args,
                                                         **kwargs))
-
         return refs if non_block else ray.get(refs)
 
     def submit(self, request: GenerationRequest) -> GenerationResult:
@@ -224,11 +238,14 @@ class RayExecutor(GenerationExecutor):
         self.workers = None
         if hasattr(self,
                    "placement_group") and self.placement_group is not None:
+            # Only remove placement group if Ray is still initialized
+            # to avoid triggering auto_init_ray() during program exit
+            if ray.is_initialized():
                 ray.util.remove_placement_group(self.placement_group)
             self.placement_group = None
         self.bundle_indices = None
 
-        if self.has_start_local_cluser:
+        if self.has_start_local_cluser and ray.is_initialized():
             logger.debug("Shutting down Ray cluster")
             ray.shutdown()
 

tensorrt_llm/executor/result.py

@@ -16,6 +16,7 @@ try:
 except ModuleNotFoundError:
     from tensorrt_llm import ray_stub as ray
 
+from .._ray_utils import unwrap_ray_errors
 from .._utils import mpi_disabled, nvtx_range_debug
 from ..bindings import executor as tllm
 from ..disaggregated_params import DisaggregatedParams
@@ -274,7 +275,7 @@ class GenerationResultBase:
             else:
                 self.queue = ray_queue
                 self.aqueue = None
-
+            with unwrap_ray_errors():
                 ray.get(self.queue.register.remote(id))
         else:
             if has_event_loop():
@@ -735,6 +736,7 @@ class GenerationResult(GenerationResultBase):
 
     def _result_step(self, timeout: Optional[float] = None):
         if mpi_disabled():
+            with unwrap_ray_errors():
                 response = ray.get(self.queue.get.remote(self.request_id))
             response = self._handle_ray_response(response)
         else:

tensorrt_llm/ray_stub.py

@@ -12,11 +12,11 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
-import functools
+from functools import wraps as _wraps
 
-from tensorrt_llm._utils import mpi_disabled
+from tensorrt_llm._utils import mpi_disabled as _mpi_disabled
 
-if mpi_disabled():
+if _mpi_disabled():
     raise RuntimeError(
         "Ray requested (TLLM_DISABLE_MPI=1), but not installed. Please install Ray."
     )
@@ -27,10 +27,11 @@ def remote(*args, **kwargs):
     def decorator(func):
         # Returns a function that always raises.
         # Decorated class depends on ray, but ray is not installed.
-        @functools.wraps(func)
+        @_wraps(func)
         def stub_checker(*_, **__):
             raise RuntimeError(
-                "Ray not installed, cannot use Ray based feature.")
+                f'Ray not installed, so the remote function / actor "{func.__name__}" is not available.'
+            )
 
         return stub_checker
 
@@ -38,3 +39,9 @@ def remote(*args, **kwargs):
         return decorator(args[0])
 
     return decorator
+
+
+def __getattr__(name):
+    raise RuntimeError(
+        f'Ray not installed, so "ray.{name}" is unavailable. Please install Ray.'
+    )

tests/unittest/llmapi/test_llm_pytorch.py

@@ -414,7 +414,6 @@ def test_llama_7b_multi_lora_evict_and_reload_evicted_adapters_in_cpu_and_gpu_ca
         repeats_per_call=1)
 
 
-@skip_ray
 @skip_gpu_memory_less_than_40gb
 def test_llama_7b_peft_cache_config_affects_peft_cache_size():
     """Tests that LLM arg of peft_cache_config affects the peft cache sizes.

tests/unittest/utils/util.py

@@ -19,7 +19,8 @@ except ImportError:
 from parameterized import parameterized
 
 import tensorrt_llm
-from tensorrt_llm._utils import torch_dtype_to_trt, trt_dtype_to_torch
+from tensorrt_llm._utils import (mpi_disabled, torch_dtype_to_trt,
+                                 trt_dtype_to_torch)
 from tensorrt_llm.llmapi.utils import get_total_gpu_memory
 from tensorrt_llm.plugin.plugin import ContextFMHAType
 from tensorrt_llm.quantization import QuantMode
@@ -449,5 +450,4 @@ def check_accuracy(a, b, atol, rtol, percent):
 
 
 skip_ray = pytest.mark.skipif(
-    os.environ.get("TLLM_DISABLE_MPI") == "1",
-    reason="This test is skipped for Ray orchestrator.")
+    mpi_disabled(), reason="This test is skipped for Ray orchestrator.")