fix: use asyncio.Event() for thread-safe initialization state

openhands-agent-server/openhands/agent_server/server_details_router.py

@@ -1,3 +1,4 @@
+import asyncio
 import os
 import sys
 import time
@@ -10,7 +11,7 @@
 server_details_router = APIRouter(prefix="", tags=["Server Details"])
 _start_time = time.time()
 _last_event_time = time.time()
-_initialization_complete = False
+_initialization_complete = asyncio.Event()
 
 
 def _package_version(dist_name: str) -> str:
@@ -60,8 +61,7 @@ def mark_initialization_complete() -> None:
     have finished initializing. Until this is called, the /ready endpoint will
     return 503 Service Unavailable.
     """
-    global _initialization_complete
-    _initialization_complete = True
+    _initialization_complete.set()
 
 
 @server_details_router.get("/alive")
@@ -83,7 +83,7 @@ async def ready(response: Response) -> dict[str, str]:
     This endpoint should be used by Kubernetes readiness probes to determine
     when the pod is ready to receive traffic. Returns 503 during initialization.
     """
-    if _initialization_complete:
+    if _initialization_complete.is_set():
         return {"status": "ready"}
     else:
         response.status_code = 503

tests/agent_server/test_server_details_router.py

@@ -0,0 +1,50 @@
+"""Tests for the server details router, including the /ready endpoint."""
+
+import asyncio
+
+import pytest
+from fastapi.testclient import TestClient
+
+import openhands.agent_server.server_details_router as sdr
+from openhands.agent_server.api import create_app
+from openhands.agent_server.config import Config
+
+
+@pytest.fixture(autouse=True)
+def reset_initialization_state():
+    """Reset the asyncio.Event between tests to avoid state leakage."""
+    sdr._initialization_complete = asyncio.Event()
+    yield
+    sdr._initialization_complete = asyncio.Event()
+
+
+@pytest.fixture
+def client():
+    app = create_app(Config(static_files_path=None))
+    return TestClient(app)
+
+
+def test_ready_returns_503_before_init(client):
+    """The /ready endpoint should return 503 while initialization is not complete."""
+    response = client.get("/ready")
+    assert response.status_code == 503
+    assert response.json()["status"] == "initializing"
+
+
+def test_ready_returns_200_after_init(client):
+    """The /ready endpoint should return 200 after mark_initialization_complete()."""
+    sdr.mark_initialization_complete()
+    response = client.get("/ready")
+    assert response.status_code == 200
+    assert response.json()["status"] == "ready"
+
+
+def test_ready_resets_after_new_event(client):
+    """After resetting the event, /ready should return 503 again."""
+    sdr.mark_initialization_complete()
+    assert client.get("/ready").status_code == 200
+
+    # Simulate a reset (e.g. for testing)
+    sdr._initialization_complete = asyncio.Event()
+    response = client.get("/ready")
+    assert response.status_code == 503