[Feature] Failure message shows more details (#2961)

Signed-off-by: wuhang <wuhang6@huawei.com>
This commit is contained in:
wuhang authored and GitHub committed 2026-04-23 04:32:43 +08:00
1 parent e3b0afbb97
commit fd9a0ee27f
6 files changed
+205 -38

No files matched your search

@@ -18,6 +18,7 @@ from pydantic import ValidationError
from pytest_mock import MockerFixture
from vllm.entrypoints.openai.engine.protocol import ErrorInfo, ErrorResponse
from vllm_omni.entrypoints.omni_base import OmniEngineDeadError
from vllm_omni.entrypoints.openai import api_server as api_server_module
from vllm_omni.entrypoints.openai.audio_utils_mixin import AudioMixin
from vllm_omni.entrypoints.openai.protocol.audio import (
@@ -1707,6 +1708,77 @@ def test_api_server_create_speech_wraps_error_response_status(mocker: MockerFixt
assert response.status_code == 400
def test_api_server_create_speech_engine_error_response_includes_request_and_stage_id(mocker: MockerFixture):
handler = mocker.MagicMock()
handler.create_speech = mocker.AsyncMock(
side_effect=OmniEngineDeadError(
"engine dead",
error_stage_id=1,
)
)
terminate_mock = mocker.patch.object(api_server_module, "terminate_if_errored")
app = FastAPI()
app.state.args = SimpleNamespace(log_error_stack=False)
app.state.openai_serving_speech = handler
app.state.engine_client = SimpleNamespace(
engine=SimpleNamespace(is_alive=lambda: False),
errored=True,
)
app.state.server = SimpleNamespace()
scope = {
"type": "http",
"app": app,
"method": "POST",
"path": "/v1/audio/speech",
"headers": [],
"query_string": b"",
"client": ("127.0.0.1", 12345),
"server": ("testserver", 80),
"scheme": "http",
}
raw_request = Request(scope)
raw_request.state.request_metadata = SimpleNamespace(request_id="speech-req-1")
request = OpenAICreateSpeechRequest(input="Hello")
response = asyncio.run(api_server_module.create_speech(request, raw_request))
assert isinstance(response, JSONResponse)
assert response.status_code == 500
assert response.body.decode("utf-8") == (
'{"error":{"message":"engine dead","type":"InternalServerError","param":null,'
'"code":500,"request_id":"speech-req-1","error_stage_id":1}}'
)
terminate_mock.assert_called_once()
def test_omni_engine_error_handler_includes_request_and_stage_id(mocker: MockerFixture):
app = FastAPI()
app.state.args = SimpleNamespace(log_error_stack=False)
app.state.engine_client = SimpleNamespace(
engine=SimpleNamespace(is_alive=lambda: False),
errored=True,
)
app.state.server = SimpleNamespace()
terminate_mock = mocker.patch.object(api_server_module, "terminate_if_errored")
api_server_module._register_omni_exception_handlers(app)
@app.get("/boom")
async def boom(request: Request):
request.state.request_metadata = SimpleNamespace(request_id="speech-req-1")
exc = OmniEngineDeadError("engine dead", error_stage_id=1)
raise exc
response = TestClient(app).get("/boom")
assert response.status_code == 500
assert response.json()["error"]["request_id"] == "speech-req-1"
assert response.json()["error"]["error_stage_id"] == 1
terminate_mock.assert_called_once()
class TestWAVHeaderGeneration:
"""Unit tests for WAV header generation with placeholder values."""
+20 -3
View File
@@ -14,6 +14,7 @@ from vllm.v1.engine.exceptions import EngineDeadError, EngineGenerateError
from vllm_omni.entrypoints.async_omni import AsyncOmni
from vllm_omni.entrypoints.omni import Omni
from vllm_omni.entrypoints.omni_base import OmniEngineDeadError
pytestmark = [pytest.mark.core_model, pytest.mark.cpu]
@@ -335,6 +336,18 @@ def _enqueue_error_message(engine: FakeAsyncOmniEngine, msg: dict[str, Any]) ->
)
def _enqueue_fatal_error_message(engine: FakeAsyncOmniEngine, msg: dict[str, Any]) -> None:
engine.output_q.put_nowait(
{
"type": "error",
"fatal": True,
"request_id": msg["request_id"],
"stage_id": 2,
"error": "engine dead",
}
)
@pytest.mark.asyncio
async def test_get_supported_tasks_returns_engine_supported_tasks():
omni = object.__new__(AsyncOmni)
@@ -546,18 +559,22 @@ async def test_async_omni_abort_forwards_to_engine(monkeypatch: pytest.MonkeyPat
@pytest.mark.asyncio
async def test_async_omni_propagates_engine_error(monkeypatch: pytest.MonkeyPatch):
engine = FakeAsyncOmniEngine(stage_metadata=THREE_STAGE_META, on_add_request=_enqueue_error_message)
async def test_async_omni_propagates_fatal_error_context(monkeypatch: pytest.MonkeyPatch):
engine = FakeAsyncOmniEngine(stage_metadata=THREE_STAGE_META, on_add_request=_enqueue_fatal_error_message)
_patch_engine(monkeypatch, engine)
app = AsyncOmni("dummy-model")
try:
with pytest.raises(RuntimeError, match="engine boom"):
with pytest.raises(EngineDeadError, match="engine dead") as exc_info:
async for _ in app.generate(prompt="hello", request_id="req-1"):
pass
finally:
app.shutdown()
assert isinstance(exc_info.value, OmniEngineDeadError)
assert str(exc_info.value) == "engine dead"
assert getattr(exc_info.value, "error_stage_id") == 2
def test_omni_generate_py_generator_yields_final_outputs_for_each_request(monkeypatch: pytest.MonkeyPatch):
sampling_params = [SamplingParams(max_tokens=8) for _ in range(3)]