rfcs VS nixops

https://www.tweag.io/blog/2020-09-10-nix-cas/ https://github.com/NixOS/rfcs/pull/62 https://github.com/NixOS/rfcs/pull/17 https://github.com/tweag/rfcs/blob/cas-rfc/rfcs/0062-content-addressed-paths.md https://github.com/wmertens/rfcs/blob/master/rfcs/0017-intensional-store.md https://discourse.nixos.org/t/content-addressed-nix-call-for-testers/12881

Content-addressable-storepath Derivations (CAS): Not something that users will mostly interact with directly, but one of the "pitfalls" of nix right now is that updates to "fundamental packages" like glibc, will cause most packages to be re-built. CAS derivation may help significantly in this regards, because there may be a point in which the changes don't have meaningful impact on a given down stream dependencies (for example, a pure python package); the CAS derivation may get to a point where it can start "re-using" existing builds after it determined that the contents exactly match. More info: https://github.com/tweag/rfcs/blob/cas-rfc/rfcs/0062-content-addressed-paths.md

As far as I know, it’s still about [0]. I’ve had a better experience with deploy-rs though [1] - or even just using nixos-rebuild to target the remote machine.

[0] - https://github.com/NixOS/nixops

For example Nix can already replace ansible, packer, cloudformation[1], dockerfiles.

Nix can also ship the nixpkgs as an oci image (e.g. docker image), vm image, iso, or if you're able to: as a nixos configuration. Tools like nixops can allow you to deploy many machines and have their behavior exactly specified, and the configuration can be version controlled. NixOS configuration can be thought of as congruent configuration management, where many other tools give you many less guarantees about configuration drift and reproducibility.

There are 4 tools I'm taking into consideration right now, but every suggestion is welcome: 1. deploy-rs - I don't know anything about it, heard about it like a day or two ago 2. NixOps - the official one, I don't know what to think, but I have concerns about Flakes compatibility 3. morph - I understand this as "NixOps, but better", no more toughs. 4. colmena - seems to be pretty straightforward with quite nice docs

I'm still relatively new to NixOS, having switched all my personal systems over to it this spring/summer. I don't have a detailed answer to your question, but I believe NixOPs is the canonical way to do what you're describing in production/at scale:

https://github.com/NixOS/nixops

https://nixos.org/nixops/manual/

Nixops is specifically made for purposes like yours.

Well, everyone of course! But especially devops, developers, power-users, and ricer folks. Due to the declarative and purity aspect of nixpkgs, all builds and configurations can be version controlled, cached and shared. NixOS can easily be extended to produce docker images, vm images, or even distributed deployments. You can also write reproducible multi-node integration tests. Tinkerers! Love playing around with the latest desktop manager or modifying builds? Nixpkgs allows you to modify any package you wish to, locally! Nixpkgs is actually a source distribution but its guarantees around purity and reproducibility are so strong that you can get a binary cache "for free".

You could use NixOps[0] for Nix but I'm not sure you can directly compare Terraform and Guix/Nix? My set up involves Terraform for infrastructure and Nix for provisioning, and it's working for me so far.

[0] https://github.com/NixOS/nixops

I'll have to read more about NixOps though, I had kind of forgotten that it existed!

Kind of off topic, but I would love to have NixOps (https://github.com/NixOS/nixops) as an abstraction layer for every type of cloud service, and not just virtual machines (e.g. queues, object storages, etc).

There is Terraform and Ansible, of course, but Nix seems like it could combine the strengths of both of them.