NAS Migration

A long time, I have described my home network, I have worked a lot on my homelab (Barracuda), and, in late December, I have decided to replace it, and my NAS.

I have replaced my 10 years old Netgear ReadyNAS 104 NAS Hannibal, and my 2 years old MINIS FORUM GK41 homelab Barracuda, by a Terramaster F4-424 Pro NAS/homelab Hannibal (yes, I have reused the name, yes, I had issue doing so).

It was motivated by many elements:

• Hannibal was really noisy during snapshots (every night)
• Barracuda was overload: the CPU was at 25% usage as a baseline (going to 100% during peaks, and was not responding), I had a 40% RAM usage and 25% swap usage
• The performances were bad: all my services were on Barracuda and the storage on Hannibal, which made their use really inefficient too
• There were too many moving parts, from time to time, I had network/power-supply issues, which forced a restart, many of the services were not able to start because the network were down, or Hannibal unreachable

I have picked the Terramaster F4-424 Pro for the following reasons:

• CPU (x5 compared to Barracuda), RAM (x4 compared to Barracuda), and 4 Bays (the same as Hannibal)
• Easiness to change the OS and plug storage (HDD & SSD)
• A solid case

So far, there's only one issue: officially, there are two USB 3.2 port, BUT, one is used by a dongle for the Terramaster OS (TOS), in order to put your own, you should unplug it, but, it is positioned such as, you can unplug it, but it stays in the port, consequently, you have only one port left.

Doing so, prevents you to have your USB dongle with the GNU/Linux Live distribution, and a keyboard to do installation instructions.

So, I have crafted a NixOS ISO image which sets up the network connection and an OpenSSH server, so I can make the actions from Looping (my desktop PC):

installer-iso = inputs.nixpkgs.lib.nixosSystem {
  specialArgs = { inherit inputs; };
  modules = [
    ({ pkgs, lib, modulesPath, ... }: {
      imports = [ "${modulesPath}/installer/cd-dvd/installation-cd-minimal.nix" ];
      nixpkgs.hostPlatform = "x86_64-linux";

      services.openssh = {
        enable = true;
        openFirewall = true;
      };


      environment.interactiveShellInit = " sleep 15 && ip addr";

      # networking.interfaces = {
      #   enp1s0.useDHCP = true;
      #   enp2s0.useDHCP = true;
      # };

      networking.firewall.enable = false;

      users.users = {
        root.initialPassword = "root";
        root.initialHashedPassword = lib.mkForce null;
        root.openssh.authorizedKeys.keys = [ "..." ];
        nixos.initialPassword = "nixos";
        nixos.initialHashedPassword = lib.mkForce null;
        nixos.openssh.authorizedKeys.keys = [ "..." ];
    };
    })
  ];
};

It worked so far, but I have encountered another issue: this NixOS was built upon nixos-unstable (nixos-25.05 preview), while Barracuda was on an older version nixos-unstable (nixos-24.05 preview), I have mechanically updated my flake.lock, which led to two issues:

• Some options were removed (I have to search extensively to find a migration process)
• Some of my services used (silently) invalid version

Let me elaborate on the last one: I was applying my Barracuda NixOS module by NixOS module. Each time I was adding one in my import, making the options' migration, deploying, if it was a successful deployment, I started again.

At some point, I checked my monitoring (one of the first module I have integrated), and it was saying "CPU used at 60%, RAM used at 40%". It was not what I was expecting, given that it was my Barracuda statistics, and Hannibal being 4-5 times more performant.

Actually, few of my services were misconfigured (the configuration changed in the new version), or using the wrong package (so I had services constantly restarting).

Then, I have added few more services previously handled by my NAS (especially the storage, which now is handled by ZFS).

I have checked my monitoring again: "CPU used at 6%, RAM used at 25%".

Great so far.

I have learned few things doing this migration:

• Put everything possible in a declarative form: along the way, I have done few manual operations on Barracuda, documented nowhere, which made me waste a lot of time
• Don't reuse existing name, reusing Hannibal without transition period was quite confusing, on my Barracuda and Looping configuration
• Always have a free storage big enough for all your data: I have choosen to reuse my Hannibal HDDs directly, sadly, I didn't have and free storage big enough (I'll write on my backup strategy in few weeks)
• Globally, things went well, it took me 3 long days and two nights for the whole process, while I've put months building my homelab