d8b059813a
* fix(startup): timeout-bound MCP session-manager start to stop M1 hang (#632) A reporter's faulthandler thread dump showed the asyncio loop alive but the lifespan suspended at an await with an idle pool worker + a leaked semaphore — the MCP Streamable-HTTP session manager hanging on its anyio task group during startup (Apple-Silicon M1). Because `enter_async_context(_sm.run())` is awaited before yield, the hang meant 'Application startup complete' never fired and the backend was unreachable with no error — a P0 (default feature dead on a platform). The MCP layer is explicitly best-effort, but the old guard only caught exceptions, not hangs. Bound the start with asyncio.wait_for (OMNIVOICE_MCP_START_TIMEOUT_S, default 30s): a hang → logged warning + backend serves without MCP. Extracted _enter_mcp_session_manager + _mcp_start_timeout_s; 4 regression tests (hang→False fast, healthy→True, None→noop, env override). No version bump. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(startup): run MCP in its own task (anyio task-affinity) — fix CI cancel-scope error The first attempt wrapped enter_async_context in wait_for, which entered the MCP anyio task group in a throwaway sub-task while the AsyncExitStack exited it on the lifespan task → 'Attempted to exit cancel scope in a different task' (caught by test_coverage_critic's real backend boot). Correct fix: _serve_mcp owns the full enter→exit in ONE task; _start_mcp_session_manager only waits (with timeout) on a ready Event. A hang still can't block startup, and enter/exit share a task. Shutdown signals stop + bounded-awaits the task. Tests updated (5; incl broken- manager case). test_coverage_critic now boots+shuts down clean. --------- Co-authored-by: mergetest <test@local> Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
84 lines
2.7 KiB
Python
84 lines
2.7 KiB
Python
"""MCP session-manager start must never wedge backend startup (#632).
|
|
|
|
On Apple-Silicon M1 the FastMCP Streamable-HTTP session manager could *hang* on
|
|
its anyio task group during lifespan startup. Because that enter was awaited
|
|
before `yield`, the hang meant "Application startup complete" never fired and the
|
|
whole backend was unreachable with no error. MCP now runs in its own task (it
|
|
owns the anyio enter→exit itself — task-affinity) and startup only *optionally*
|
|
waits, with a timeout, on a ready signal: a hang → a logged warning + a backend
|
|
that still serves without MCP.
|
|
"""
|
|
import asyncio
|
|
import os
|
|
import sys
|
|
|
|
sys.path.insert(0, os.path.join(os.path.dirname(os.path.dirname(os.path.abspath(__file__))), "backend"))
|
|
|
|
from main import _start_mcp_session_manager, _mcp_start_timeout_s # noqa: E402
|
|
|
|
|
|
class _CM:
|
|
def __init__(self, hang, raise_on_enter=False):
|
|
self.hang = hang
|
|
self.raise_on_enter = raise_on_enter
|
|
|
|
async def __aenter__(self):
|
|
if self.raise_on_enter:
|
|
raise RuntimeError("boom")
|
|
if self.hang:
|
|
await asyncio.sleep(60) # never completes within the test timeout
|
|
return self
|
|
|
|
async def __aexit__(self, *a):
|
|
return False
|
|
|
|
|
|
class _SM:
|
|
def __init__(self, hang=False, raise_on_enter=False):
|
|
self._cm = _CM(hang, raise_on_enter)
|
|
|
|
def run(self):
|
|
return self._cm
|
|
|
|
|
|
def _drive(sm, timeout):
|
|
"""Run _start_mcp_session_manager, then clean up the task; return `mounted`."""
|
|
async def go():
|
|
task, stop, mounted = await _start_mcp_session_manager(sm, timeout=timeout)
|
|
stop.set()
|
|
if task is not None:
|
|
task.cancel()
|
|
try:
|
|
await task
|
|
except BaseException:
|
|
pass
|
|
return mounted
|
|
return asyncio.run(go())
|
|
|
|
|
|
def test_hang_does_not_block_startup():
|
|
# The crux: a hanging manager returns fast with mounted=False (no raise).
|
|
assert _drive(_SM(hang=True), 0.2) is False
|
|
|
|
|
|
def test_healthy_manager_mounts():
|
|
assert _drive(_SM(hang=False), 5.0) is True
|
|
|
|
|
|
def test_broken_manager_is_not_mounted():
|
|
# An exception during enter → not mounted, startup still proceeds.
|
|
assert _drive(_SM(raise_on_enter=True), 5.0) is False
|
|
|
|
|
|
def test_none_manager_is_noop():
|
|
assert _drive(None, 5.0) is False
|
|
|
|
|
|
def test_timeout_env_override(monkeypatch):
|
|
monkeypatch.setenv("OMNIVOICE_MCP_START_TIMEOUT_S", "12.5")
|
|
assert _mcp_start_timeout_s() == 12.5
|
|
monkeypatch.delenv("OMNIVOICE_MCP_START_TIMEOUT_S", raising=False)
|
|
assert _mcp_start_timeout_s() == 30.0
|
|
monkeypatch.setenv("OMNIVOICE_MCP_START_TIMEOUT_S", "garbage")
|
|
assert _mcp_start_timeout_s() == 30.0 # invalid → safe default
|