NixOS

NixOS takes infrastructure as code further than any configuration management tool: the operating system itself is the output of a declarative expression. The packages installed, the services running, the users, the firewall rules and the kernel parameters are all defined in a file, and applying that file produces exactly that system.

The approach came out of Eelco Dolstra's doctoral research at Utrecht University, and the NixOS Foundation that stewards it is Dutch.

Rollback That Actually Works

Every system change creates a new generation, and the previous one remains bootable. If an upgrade breaks something, you reboot into the previous generation and you are back exactly where you were — not approximately, exactly. Anyone who has watched a package upgrade leave a server in an inconsistent half-state understands why this matters.

Because changes are atomic, there is no window in which the system is partially updated.

Reproducible Environments

Nix development shells give each project its own precisely pinned toolchain without polluting the machine or requiring containers. Two developers on different machines, and CI, get byte-identical environments from the same expression — which removes a large category of "works on my machine" problems.

For long-lived systems, being able to rebuild the exact environment from five years ago is a genuine capability rather than an aspiration.

As Configuration Management

For fleets, NixOS replaces the role Ansible or Puppet plays, but with a stronger guarantee: configuration management tools converge towards a desired state and can drift, whereas a NixOS generation is the state. There is no partial application to reconcile.

Limitations to Consider

The learning curve is the steepest of anything in this category. The Nix language is unusual, documentation has historically been uneven, and debugging a failing build often means reading other people's expressions. Software that expects a conventional filesystem layout frequently needs packaging work. Adopting NixOS is a strategic decision for a team, not a tool you pick up in an afternoon.

Who Should Use NixOS

Teams that manage many machines and value reproducibility and reliable rollback highly enough to invest in the learning curve — research computing, build infrastructure, embedded fleets and organisations with long-lived systems that must be rebuildable years later.