Better real-world testing

[mkg20001]

After #1213 there's now https://github.com/NixOS/nixops/issues/1005

The hetzner module seems to be broken at pretty much every deployment with another error message coming along to annoy me.

Would it be possible to put an end to this madness?

[mkg20001]

I've just got an idea: There could be a hetzner machine that's specifically used for testing nixops on it's own account that would be simply re-deployed at every test.

[paluh]

@mkg20001 I fully understand your proposition as I've spent last few days learning (I'm a slow learner :-P) and playing with nixops and hetzner provisioning. Maybe we should move this discussion to the NixOS/nixops-hetzner repo.

P.S.
I was able to finally get working hetzner deployment.

I'm using latest master branch of nixops and pinning some nixpkgs-19.0.9 version (to have really reproducible deployment) in this shell.nix:

let
  hostPkgs = import <nixpkgs> {};
  # nixos-19.0.9 - v2020-01-13
  pkgsSrc = hostPkgs.fetchFromGitHub {
    owner = "NixOS";
    repo = "nixpkgs-channels";
    rev = "09aa1b23bb5f04dfc0ac306a379a464584fc8de7";
    sha256 = "1byv12ayy7k6p64jgm3054xif8yvp388sr3lycd4ax01w9qpc1ry";
  };
  pkgs = import pkgsSrc {
    config.allowUnfree = true;

    overlays = [
      (self: super: {
        nixops =
          let
            src = pkgs.fetchFromGitHub {
              owner = "NixOS";
              repo = "nixops";
              rev = "4cfb70513bad149183adc3ac741c176d83b0e9d5";
              sha256 = "0irf9wha2rxla6z7mywj5z29bvjbpwlxqj2s29ygsbhp6hnlbzzz";
            };
          in
            (import "${src}/release.nix" { p = (p: [ (p.hetzner) ]); }).build.x86_64-linux;
      })
    ];
  };
in
  pkgs.stdenv.mkDerivation {
    name = "nixops-trivial";
    NIX_PATH="nixpkgs=${pkgs.path}";
    buildInputs = [ pkgs.nixops ];
  }

Minimal empty logical machine definition - trivial.nix:

{ ketosis =
    { config, pkgs, ... }: {};
}

I wanted to use raid with level=0 but I wasn't able to boot machine again. Debugging boot issue was outside of my current scope so I've decided to use lvm instead - trivial-hetzner.nix:

{ ketosis = {
    deployment.targetEnv = "hetzner";
    deployment.hetzner.mainIPv4 = "*.*.*.*";

    deployment.hetzner.createSubAccount = false;

    deployment.hetzner.partitions = ''
      clearpart --all --initlabel --drives=nvme1n1,nvme0n1

      part pv.01 --label=root-part1 --grow --ondisk=nvme0n1
      part pv.02 --label=root-part2 --grow --ondisk=nvme1n1

      volgroup vg0 pv.01 pv.02
      logvol swap --vgname=vg0 --recommended --fstype swap --label=swap --name=swap
      logvol /    --vgname=vg0 --size=1800000 --fstype ext4 --label=root --name=root
    '';
  };
}

At the end I've got some errors from nixops command and its reboot didn't finish but when I had reseted this machine from my hetzner robot panel everything worked fine.

updating GRUB 2 menu...                                                                                                                                       
installing the GRUB 2 boot loader on /dev/nvme0n1...                                                                                                          
Installing for i386-pc platform.                                                                                                                              
You have a memory leak (not released memory pool):                                                                                                            
[0x56de50] dtree 

In order to test this configuration you should:

• Check and setup your device labels (in standard raid level=1 partitioning setup there are by default two disks in use sda and sdb) in partitioning section.

• Export HETZNER_ROBOT_USER and HETZNER_ROBOT_PASS into your env.

• Set up your machine IP address in trivial-hetzner.nix deployment.hetzner.mainIPv4 because I've marked it out.

• It is important to use deployment.hetzner.createSubAccount = false; as it is because there is an open issue related to this feature.

[paluh]

P.S.II.
Somewhat related: NixOS/nixops-hetzner#6

[mkg20001]

A bit related, since it could make testing better and automate reproducing the issues for fixing them: NixOS/nixpkgs#2079