Interesting writeup. I'm curious about the resource usage of the Lemmy backend and frontend deployments. Do you have any insights on the resource utilization of these deployments?
While setting up pict-rs on my cluster I noticed a pict-rs environment variable to convert all uploaded pictures to the png format. That may also be activated. I guess that makes de-duplication way easier.
In my opinion restarting with rust is currently not the best idea. There are a lot of more polished tools available for other languages. But if you are sure about using rust, you can check out arewegameyet. It lists a lot of tools and crates to start game development with rust. I tried out bevy once, but couldn't really figure it out in a time I was happy with. Doing stuff like this in rust includes a lot of reading (auto generated) documentations and just trying stuff out again, again and again.
I created a helm chart to help setting up Lemmy on kubernetes. I am already using it successfully for my instance.
You can find the charts and a short documentation here:
I deployed Lemmy on my kubernetes cluster. You can find a short guide and my helm charts here:
https://lemmy.help/post/6845
https://github.com/grouvie/lemmy_help
I am using argoCD to manage my charts. Feel free to contact me if you have any questions.