Managing small nix fleet

NixOS comes with a huge set of options to configure the system declaratively. It covers all base needs — bootloader, fstab, display managers, Mesa drivers, and everything else required for a fully functional system that meets daily needs. No magic happens here: if you can't make your workflow fit a regular Ubuntu setup, Nix won’t save you.

Additionally, NixOS offers versatile configurations for different services. Be it Nginx, PostgreSQL, Redis, *arr, Wakapi, Docker, Gitea, Grafana, Prometheus, or Nextcloud — you name it. Support quality varies across services. In the best case, you get rich, typed configurations; on average, you get a predefined systemd service and the ability to inject your configuration as a Nix-lang key-value map. Worst case? You still have Docker, which Nix supports quite well — allowing you to define containers similarly to Docker Compose but with Nix syntax.

Community involvement is crucial for any large project, and the Nix community delivers great tools.

I use not so much things in my daily workflow, but here is a list:

• stylix awesome package for applying system wide styling. Like you don't need to go through all supported apps and change color scheme. Pick one of base16 and you will get theme almost everywhere.
• Home Manage Simple explanation but it do not explain depth of module: "It is like nixos but for your user home folder". Manage your user personal belongings in a nix way. Install apps only for this user, configure Sway, Firefox, git, gpg, ssh, shell.
• sops-nix Provide secrets on machine or user level. It may work for you too.
• nur-expressions I use it to manage addons for my firefox.
• nixos-hardware Self explanatory. Bits for fast prototyping hardware configuration.
• deploy-rs Spread your flakes on your fleet. Designed for remote provision, but works for local too.
• nixos-rk3588 Usefull set of modules for RK3588 SBCs. Sadly it is archived now.
• nixos-infect Infect machine from your cloud provider with Nixos.

For me it is a huge deal. I use it both at work and for personal projects. Combining flakes, direnv, and nix-direnv gives me ready-to-use development environments instantly. If you’re curious, read Amos's excellent series: https://fasterthanli.me/series/building-a-rust-service-with-nix/part-10

Nix makes it simple to package apps if they use common build systems like Meson. I build a couple of 3rd party forks of mpp, ffmpeg and libmali myself. Works fine. Also you can modify any package from nixpkgs with reach overlay system.

I’m used to rolling-release distros, so I understand "stability" in my own way: I don’t update systems during the workweek, only on Friday evenings. Tbh, this stability is good enough. Paired with generations and away to rollback on boot you get a robust system.

However, configuration stability is a mixed bag. Even on release branches of nixpkgs, I occasionally encounter breaking changes in configuration options over short timeframes. Community-driven modules are less stable.

I don’t blame anyone—this is just how decentralized projects work. Pinning module versions and updating them sparingly helps manage this pain point.

Welp, it’s too late to blame systemd for everything, but in NixOS, services, hooks, and one-time jobs all rely exclusively on systemd units. That’s it. I’m not a fan.

Nix is a pure, functional, lazy, declarative, and reproducible programming language.

Just keep in mind that it is not general purpose programming language. If you used to something mainstream you will suffer. And LSP would not make your life easier.

The first time you hit infinite recursion encountered, you’ll start appreciating Rust’s compiler errors. Nix being pure and lazy means you can’t simply scatter logs for debugging without jumping through hoops.