app: 600req/s
app + nginx: 20req/s -> 400req/s
that's going to feel great once deployed on m.s
turns out we may have no keepalive between proxies
@CobaltVelvet I once had a job where I spent a week fixing some performance bottlenecks, then another two to convince management the 400% performance bump wasn't an error 😅
@CobaltVelvet *poticâlin*
the benchmark was SO broken
the proxy was actually working perfectly fine the whole time
forgot the "X-Forwarded-Proto: https" so my non-proxied measurement was just measuring a redirection to https 🤦♀️
if it looks too good the benchmark is probably broken
it was broken