sozu VS tokio-reuseport

This is interesting, thank you.

I really should learn from BEAM and the OTP and learn Erlang. I get the feeling it's super robust and reliable and low maintenance. I wrote a userspace multithreaded scheduler which distributes N lightweight threads to M kernel threads.

https://github.com/samsquire/preemptible-thread

I recently wrote a JIT compiler and got lazy compilation of machine code working and I'm nowhere near beginning optimisation

https://github.com/samsquire/compiler

How do you write robust software, that doesn't crash when something unexpected goes on?

I looked at sozo https://github.com/sozu-proxy/sozu

and I'm thinking how to create something that just stays up and running regardless.

It's AGPL licensed which for a proxy is a strange choice. They have an unanswered question for months on what it might mean: https://github.com/sozu-proxy/sozu/issues/764

Without an answer to that if you use this and you need to make any change to it (even a tiny bug fix), you're basically opening yourself up a pile of legal issues.

Sozu: Well documented, runtime configurable proxy

https://github.com/sozu-proxy/sozu https://github.com/sozu-proxy/sozu/blob/main/doc/design_motivation.md

Could sozu or rathole or leaf or exodus somehow help?

it is not sufficient: a listen socket has its own queue of new TCP connections (that were already handshaked by the kernel), so dropping the listen socket drops the queue. The right way is to start the new server, transfer the listen socket from the old server to the new one with SCM_RIGHTS, then start accepting again from the new instance. That's how it is done in the sozu HTTP proxy (which also uses SO_REUSEPORT to launch multiple work processes each with their own listeners, to improve performance and isolate failure)

BTW that traffic will be coming from our sozu load balancers, built in Rust too ;)

How does it compare to sozu?

The code linked in the article is quite explanatory - you have not a single listening socket, but per-core each.