Merge pull request 'Bound readiness probe latency' (#72) from timmy/71-bound-readiness-probe-latency into main

src/main.py

@@ -15,6 +15,7 @@ from src.views import router as frontend_router
 app = FastAPI(title="Stackchain Dashboard")
 CONTEXT_TIMEOUT_SECONDS = 5.0
 EVENT_STREAM_TIMEOUT_SECONDS = 5.0
+READINESS_TIMEOUT_SECONDS = 5.0
 FRONTEND_DIR = Path(__file__).resolve().parent.parent / "frontend"
 
 app.add_middleware(
@@ -39,15 +40,26 @@ def health() -> dict[str, str]:
 async def readiness():
     """Return readiness after verifying the configured Gitea connection."""
     try:
-        user = await current_user()
+        user = await asyncio.wait_for(
+            current_user(), timeout=READINESS_TIMEOUT_SECONDS
+        )
     except Exception as exc:
+        timed_out = isinstance(exc, TimeoutError)
+        error_message = (
+            f"Gitea readiness check timed out after {READINESS_TIMEOUT_SECONDS:g}s"
+            if timed_out
+            else str(exc)
+        )
         return JSONResponse(
             {
                 "status": "not_ready",
                 "service": "stackchain-dashboard",
-                "error": str(exc),
+                "error": error_message,
             },
             status_code=503,
+            headers={
+                "Retry-After": str(max(1, math.ceil(READINESS_TIMEOUT_SECONDS)))
+            } if timed_out else None,
         )
     return {
         "status": "ready",

tests/test_health.py

@@ -1,3 +1,6 @@
+import asyncio
+import json
+
 import pytest
 
 from src import main
@@ -37,3 +40,29 @@ async def test_readiness_endpoint_returns_503_when_gitea_is_unavailable(monkeypa
     assert response.status_code == 503
     assert b'"status":"not_ready"' in response.body
     assert b'"error":"connection refused"' in response.body
+
+
+@pytest.mark.anyio
+async def test_readiness_endpoint_times_out_and_cancels_stalled_gitea_check(monkeypatch):
+    cancelled = asyncio.Event()
+
+    async def hanging_user():
+        try:
+            await asyncio.sleep(0.05)
+            return {"login": "too-late"}
+        finally:
+            cancelled.set()
+
+    monkeypatch.setattr(main, "READINESS_TIMEOUT_SECONDS", 0.01, raising=False)
+    monkeypatch.setattr(main, "current_user", hanging_user)
+
+    response = await main.readiness()
+
+    assert response.status_code == 503
+    assert response.headers["retry-after"] == "1"
+    assert json.loads(response.body) == {
+        "status": "not_ready",
+        "service": "stackchain-dashboard",
+        "error": "Gitea readiness check timed out after 0.01s",
+    }
+    assert cancelled.is_set()