Comment on how to start with self-hosting?

• downhomechunk@midwest.social ⁨7⁩ ⁨hours⁩ ago

  Your approach works too. Something like CasaOS answers OP’s question directly. I was thinking about how I started on this journey. I wanted to play with enterprise level tools at home on repurposed e-waste. So I started with proxmox. But I also came to the table with a couple decades of Linux experience under my belt.

  Those scripts make it so easy. You can paste a command, accept defaults, watch some text scroll by and finish with instructions on how to access the tool you just installed.

  My homelab is low power as well. I’m currently running zero VMs. Everything is done with LXCs. You can run a pi hole on 512 MB RAM.

  source
• Windex007@lemmy.world ⁨2⁩ ⁨days⁩ ago

  Easily can have multiple LXCs, and being able to take snapshots for backup is probably a nice thing to have if you’re just learning.

  And if they get more hardware, moving VMs to other clustered proxmox instances is a snap.

  source

  • sugar_in_your_tea@sh.itjust.works ⁨2⁩ ⁨days⁩ ago

    If you just want LXCs, use Docker or Podman on whatever Linux distro you’re familiar with. If you get extra hardware, it’s not hard to have one be the trunk and reverse proxy to the other nodes (it’s like 5 lines of config in Caddy or HAProxy).

    If you end up wanting what Proxmox offers, it’s pretty easy to switch, but I really don’t think most people need it unless they’re going to run server grade hardware (i.e. will run multiple VMs). If you’re just running a few services, it’s overkill.

    source

    • Windex007@lemmy.world ⁨2⁩ ⁨days⁩ ago

      If you’re just running a few services, and will only ever be running a few services, I agree with you.

      The additional burden of starting with proxmox (which is really just debian) is minimal and sets you up for the inevitable deluge of additional services you’ll end up wanting to run in a way that’s extensible and trivially snapshotable.

      I was pretty bullish on “I don’t need a hypervisor” for a long time. I regret not jumping all-in on hypervisors earlier, regardless of the services I plan to run. Is the physical MACHINEs purpose to run services and be headless? Hypervisor. That is my conclusion as for what is the least work overall. I am very lazy.

      source

      • sugar_in_your_tea@sh.itjust.works ⁨2⁩ ⁨days⁩ ago

        For snapshots, you can use filesystem features, like BTRFS or ZFS snapshots. If you make sure to encapsulate everything in the container, disaster recovery is as simple as putting configs onto the new system and starting services (use specific versions to keep things reasonable.

        I think that’s also really lazy, it’s just a different type of lazy from virtualization.

        My main issue with virtualization is maintenance. Most likely, you’re using system dependencies, and if you upgrade the system, there’s a very real chance of breakage. If you use containers, you can typically upgrade the host without breaking the containers, and you can upgrade containers without touching the host. So upgrades become a lot less scary since I have pretty fine-grained control and can limit breakage to only the part I’m touching, and I get all of that with minimal resource overhead (with VMs, each VM needs the whole host base system, containers don’t).

        Obviously use what works for you, I just think it’s a bit overwhelming for a new user to jump to Proxmox vs a general purpose Linux distro.

        source