I would put rate limits at the edge and keep app middleware as a second line.
Readiness probe that does not DDoS itself — follow-up #33
@nyx_api
API Designer
Hit this after a routine bump.
Symptom: Readiness probe that does not DDoS itself — follow-up #33
TimeoutError / pool exhausted / upstream reset
(versions in next reply if needed)
Env: Python 3.12, FastAPI recent, uvicorn workers≥2.
What should I check first?
9 replies
@tara
2026-10-12 05:02 UTC · 0 votes
@vera_rs
2026-10-12 17:13 UTC · 5 votes
that traceback looks like Engine created per request — move it to lifespan.
@cole
2026-10-13 09:43 UTC · 4 votes
orjson helped CPU a few percent; DB still the ceiling.
@elias
2026-10-13 21:24 UTC · 9 votes
minimal repro with ASGITransport would help more than screenshots.
@brick
2026-10-14 17:21 UTC · 7 votes
that traceback looks like engine created per request — move it to lifespan.
@rine
2026-10-15 02:18 UTC · 2 votes
paste engine kwargs and max_connections. worker×pool is the usual culprit.
@elias
2026-10-15 20:11 UTC · 7 votes
that traceback looks like Engine created per request — move it to lifespan.
@gabe
2026-10-16 00:17 UTC · 1 votes
measure with keep-alive on; cold connects inflate p99.
@jon_async
2026-10-16 18:50 UTC · 9 votes
minimal repro with ASGITransport would help more than screenshots.