Hey @hamiltop! To clarify, enabling keep-alive on your server does not work in deployment, but it works locally? I suspect this is due to your application being behind our proxies in deployment, and the proxies don't have keep_alive enabled. Whether they should warrant further investigation, though.

That would make sense. I would strongly suggest enabling keep-alive on the proxies, both from client->proxy and proxy->server. Throughput is much better due to reduced tcp connection overhead (and may be even more important for HTTPS efficiency).

Thanks for the suggestion @hamiltop! We'll need to do some investigation to explore potential drawbacks, as well as what to set it to and where to set it, but it makes sense! If you have thoughts on that, it would be great if you could expand on it a bit, if you have the time.

For context, requests to your service first go through an AWS LB, then to a proxy in shuttle-gateway, then to a proxy in shuttle-deployer (your project container) and on to your application. You can find the code for our proxies in gateway/src/proxy.rs and deployer/src/proxy.rs if that interests you.

I dug in a bit and found that hyper-reverse-proxy removes the connection header which is valid for http 1.1. That does break keep-alive for http 1.0. Maybe http 1.0 isn't that big of a deal. It only showed up on my radar because I tried using apache bench (ab) to benchmark things and it's an http 1.0 client.