I would like to have a mechanism to set up a server automagically…

Similarly I would like to set up my user account settings (Tmux plugins, .zshrc and vim settings, etc) that I can replicate in multiple machines via a script (I have a custom script for this but I want a more solid alternative)

Thoughts on what infra-as-code solution would work best? Any similar experiences or use cases with one Thanks!
Cc @selfhost@lemmy.ml @selfhosted@lemmy.world

  • MajorHavoc@lemmy.world
    link
    fedilink
    arrow-up
    17
    arrow-down
    1
    ·
    11 months ago

    I think you’re looking for Ansible. Have fun!

    The difference between an Anible playbook and a script, is Ansible has a ‘check’, ‘change’, ‘verify’ pattern, and is declarative (meaning that once the playbook is made, it tends to keep working on future versions of Ansible.)

    • marx2k@lemmy.world
      link
      fedilink
      arrow-up
      1
      arrow-down
      2
      ·
      11 months ago

      Ehhh… as someone who does devops, you should dive into ansible core changelogs on github sometime ;)

      • MajorHavoc@lemmy.world
        link
        fedilink
        arrow-up
        2
        ·
        11 months ago

        I assume you mean to check on his often they’re is the breaking changes? :)

        Declarative style isn’t perfect, but it’s a massive improvement from straight bash scripting.

        • marx2k@lemmy.world
          link
          fedilink
          arrow-up
          2
          arrow-down
          1
          ·
          11 months ago

          💯

          We’re an ansible shop and yeah it’s better than bash scripting (where it makes sense) but ansible… man it does have some peculiarities :/

  • Atemu@lemmy.ml
    link
    fedilink
    arrow-up
    16
    arrow-down
    2
    ·
    11 months ago

    I would like to have a mechanism to set up a server automagically…

    NixOS.

    Similarly I would like to set up my user account settings

    Home-manager.

  • Voroxpete@sh.itjust.works
    link
    fedilink
    arrow-up
    12
    ·
    edit-2
    11 months ago

    “I want to automatically build VMs, networks, and other infra in the cloud using repeatable specifications.”

    Terraform

    “I want to host my own cloud (either by paying for bare metal hosting, or providing my own) that I can deploy those VMs on.”

    Openstack or Openshift

    “I want to automatically configure servers after deployment.”

    Ansible

    “I want to deploy services to those servers in a simple, repeatable fashion.”

    Docker, Podman, or Kubernetes.

      • Voroxpete@sh.itjust.works
        link
        fedilink
        arrow-up
        1
        ·
        edit-2
        11 months ago

        I completely agree. But I’m working from the assumption that if OP actually needs the ability to automatically provisions platforms, they’re probably working at a larger scale than the typical small self-hosted home server. And I like to give options.

        IMO, the only tools here that most self hosters need are Docker, and maybe Ansible, though even that is a stretch because in most cases you’re just going to have one server running all your containers and that’s it.

        It’s the whole “Cattle vs Pets” question. When you’re a typical self hoster, you’re probably better off just treating every server as a pet. But if you’re using self hosting as a way of building job skills (which is exactly how I broke into IT) then you absolutely want to start learning how to wrangle cattle.

  • sudneo@lemmy.world
    link
    fedilink
    arrow-up
    10
    ·
    11 months ago

    Ansible is definitely one way to do this. If your machines are VMs, then also building VM images with packer can be the way.

    For tmux, vim, etc. You can still use ansible or some specific tool for dotfiles, like chezmoi (there are a bunch). You can even use ansible to run chezmoi!

  • Moonrise2473@feddit.it
    link
    fedilink
    arrow-up
    4
    ·
    11 months ago

    If it’s because you set new servers very frequently: Ansible

    If it’s because moving stuff once every two years to a new server is an hassle: everything in custom docker images

    • taladar@sh.itjust.works
      link
      fedilink
      arrow-up
      1
      ·
      11 months ago

      If you don’t update your Docker images for two years or more you are going to have even more security holes than if you had it all on the host system and didn’t update that.

      • Moonrise2473@feddit.it
        link
        fedilink
        arrow-up
        1
        ·
        11 months ago

        I build the docker images every time I call docker compose, they should be updated

        There’s the huge bug that apparently they don’t want to fix, is that it caches “apt upgrade” so I have to edit that line every once in a while

        One day I will update my script to add a random commented char at the end of that line every time

  • johntash@eviltoast.org
    link
    fedilink
    English
    arrow-up
    3
    ·
    11 months ago

    You’re probably looking for some sort of configuration management tool like chef, ansible, saltstack, or puppet. If you’re not already familiar with one, ansible is pretty easy to get started with.

    If you’re also wanting something that can create the server itself, terraform is great and supports most cloud providers and supervisors.

  • bouh@lemmy.world
    link
    fedilink
    arrow-up
    1
    ·
    11 months ago

    Why is infra as a code so sought after? I feel like this is installation scripts and config like bare bones, but you need another layer to make it work on top. What am I missing?

    • taladar@sh.itjust.works
      link
      fedilink
      arrow-up
      2
      ·
      11 months ago

      Not sure why you would use it for a single server with a single admin you only install once but for multiple admins and many servers it provides repeatable results that are the same no matter who does it and it also allows you to add small settings that you would never do by hand every time you install a new machine. There is nothing worse than discovering that your dev system and your production system differ in a minor way that makes a test succeed on dev but fail on production because of something someone installed or configured manually. Well, apart from discovering that same thing happened with your 5 year old production server you are trying to reinstall after it broke.

      • bouh@lemmy.world
        link
        fedilink
        arrow-up
        2
        ·
        11 months ago

        What I mean is that repeatability can be achieved in other, simpler ways. Like a package for example.

        I feel like as technologies, ansible and docker have been spread beyond their relevant scope of usefulness. But maybe that’s me.

        I feel like ansible is a complex way of doing simple things.

        • taladar@sh.itjust.works
          link
          fedilink
          arrow-up
          2
          ·
          11 months ago

          Packages seem like a very convoluted way to achieve something like setting a host name or configuring the DNS server a system uses or the packages that are installed or which virtual hosts a web server serves and which certificates it uses to do so.

  • thirdBreakfast@lemmy.world
    link
    fedilink
    English
    arrow-up
    1
    ·
    11 months ago

    Bare metal servers, VPSs, or VM’s you host? If it’s for VM’s you host, then consider Proxmox as hypervisor and use VM templates. I’m sure old school sysops could to the same with QEMU and Virtmanager or something. But basically, I just set up a VM exactly how I like it, then convert it to a template and cookie cutter it out.

    I can sense the Nix guys shaking their heads - it’s on my list to try :- )