A home server stops being a project and starts being infrastructure the day you stop treating it like a tinkering hobby. The hardware is the easy part. The operating system you install underneath everything else is the decision that quietly locks you in for the next three to five years, and most first-time builders discover that on the weekend they try to migrate off it. This article is about why the boring choice wins, where the boring choice breaks down, and the five places even careful home server owners get bitten because they treated the base OS as an afterthought.
If you have ever spent a Saturday recovering from a Friday-night “let me just try a different hypervisor” (a hypervisor is the software layer that lets one physical machine run multiple virtual machines side by side, each thinking it owns the hardware), the rest of this article will feel familiar. If you have not, the rest of this article is what you should read before you install anything.
Why the base OS is the only decision that costs you a year
Most home server guides start with the apps. Plex for media. Home Assistant for automation. Pi-hole for ad blocking. Maybe a Minecraft server for the kids. That list is fine. It is also the wrong place to start, because every app you install sits on top of a foundation that, once chosen, is painful to change. The foundation is the operating system that runs the box, the storage layout underneath it, and the way the system handles updates. None of those decisions are visible from the app layer. All of them are expensive to redo later.
Pick the wrong hypervisor and you will spend a weekend re-installing every container and every VM (virtual machine, a software-emulated computer that runs inside your real computer as if it were a separate physical box). Pick the wrong storage layout and you will spend a month migrating terabytes of data off a dying drive, hoping you did not lose the family photos. Pick the wrong update policy and you will wake up one Sunday to a broken automation stack because the upstream package decided to remove a deprecated API (an application programming interface, the documented way one piece of software calls into another) you depended on.
The cost of getting the base OS right is roughly four hours of reading. The cost of getting it wrong is somewhere between one bad weekend and one full quarter of rebuilds. The four hours is the cheaper investment by an order of magnitude, and yet almost everyone skips it.
What “boring” actually means in this context
Boring does not mean old. Boring means the operating system you pick has been around long enough that someone has already written the recovery guide for the failure you are about to hit. Boring means the documentation is on a real domain, not a Discord pin from 2021. Boring means the upgrade path from your current version to the next version is documented and tested, not “watch this GitHub issue for the workaround.”
For most home server builders in 2026, boring is one of three choices. Proxmox, if you want a real hypervisor with a web UI and the ability to run containers and VMs on the same box. Unraid or TrueNAS Scale, if your main job is network attached storage (a big pool of drives that everyone on your home network can read and write to, often shortened to NAS). Plain Debian with Docker on top, if you already know exactly what you want and do not need a hand-holding web UI.
What boring is not, and this matters, is the latest thing on Hacker News. The latest thing on Hacker News is fun to read about and absolutely the wrong choice for the foundation of your home infrastructure. The cost of being on the bleeding edge for your base OS is that every other tutorial you read assumes a different base, every recovery guide assumes a different layout, and every update cycle breaks something the maintainer did not document.
The rule of thumb is simple. If you cannot find a recovery guide written by someone who is not the project maintainer, do not put it at the base of your stack. Pick something boring enough that strangers have written the recovery guides.
The five places boring breaks down
Boring is the right answer most of the time, but it is not a complete answer. There are five specific failure modes where even a careful home server owner gets bitten, and you should know them in advance.
First, boring does not mean hands-off. The whole point of Proxmox is that it gives you a stable foundation, but you still have to apply updates on a schedule you control. If you let the update queue sit for six months because the web UI never nags you, you will eventually hit an update that requires a manual step the silent queue did not perform. The fix is a calendar reminder, not a different base OS.
Second, boring does not mean fast. A stable Debian base is going to ship older package versions than the latest Ubuntu LTS (long-term support release, the version Ubuntu commits to updating for five years). If you need a specific modern feature in a specific app, the older package versions might not support it yet. The fix is usually a single Docker container running the modern app on top of your boring base, not a switch to a newer base.
Third, boring does not mean compatible. Unraid’s storage layout is genuinely good for media, but it makes some backup tools behave in surprising ways. TrueNAS Scale is genuinely good for ZFS (ZFS is a copy-on-write filesystem, meaning the filesystem never overwrites existing data in place but writes new copies, which makes corruption self-healing in many cases), but its container story is less mature than Proxmox’s. Proxmox is genuinely good for everything, but its storage layer assumes you understand LVM (logical volume management, a way of carving up physical disks into flexible virtual partitions) better than you might want to. Pick the boring that matches your actual main use, not the boring that sounds most flexible.
Fourth, boring does not mean documented-for-your-version. The Proxmox wiki is good, but most of the third-party tutorials assume Proxmox 7, not Proxmox 8. Debian’s wiki is good, but the debian.org instructions lag behind the package repository by a release or two. The fix is reading the changelog before you install, not after something breaks.
Fifth, boring does not mean you can ignore hardware. A boring base OS cannot save you from a dying hard drive that you never monitored, a power supply that browns out (briefly drops below the voltage your hardware needs, causing glitches rather than a clean shutdown) during a storm, or a router that quietly stops passing DHCP (the protocol your router uses to hand out IP addresses to devices on your home network) for an hour while you are asleep. The boring base gives you the tools to monitor those things. It does not monitor them for you.
The order I would install things in this weekend
Start with the base OS and nothing else. Proxmox, Unraid, TrueNAS, or Debian with Docker, depending on which boring fits your main use. Install it, update it, reboot it once, and walk away for an hour. If it survives that hour, you have a foundation. If it does not, you have learned something useful before you invested any time in apps.
Next, set up storage. If you are on Proxmox, decide whether you are using ZFS, Btrfs, or ext4 (three common filesystems, the on-disk format your OS uses to organize files; Btrfs and ZFS both support snapshots and integrity checks that ext4 does not) for the boot drive and a separate layout for the storage pool. If you are on Unraid or TrueNAS, configure the storage pool before you put a single file on it. Storage decisions are the second-most-expensive to undo after the base OS decision, and you should make them on day one even if you only have one drive in the box.
From there, set up remote access before you expose anything. Tailscale is the easiest mesh VPN (a virtual private network that makes your devices behave as if they are all on the same local network, no matter where they physically are) for beginners, free for up to a hundred devices, and it does not require opening a port on your router. Wireguard is the heavier version if you want to learn the protocol. Cloudflare Tunnel works for specific web services but is not a general VPN. Pick the right tool for the job and install it before you install any other app.
Before you add any service, write down what you did. A single markdown file you can find six months from now, with one line per service: what it is, why you installed it, and where its data lives. The most expensive part of a home server is not the hardware. It is the half-day you spend in six months trying to remember why you configured something the way you did.
Then, and only then, add one service per session. Plex this week. Home Assistant next week. Pi-hole the week after. Slow is smooth, smooth is fast, and your future self will not have to migrate off a bad decision in a panic because you gave yourself room to learn the base before you stacked anything on top.
- Boring base OS, installed first, updated once, then walked away from for an hour
- Storage layout decided on day one even if you only have one drive
- Remote access set up via Tailscale before any service touches the internet
- A single markdown notes file with one line per service, written before adding apps
- One service per session, with a real week between additions
Trade-offs
The boring choice costs you time upfront that the exciting choice lets you skip. Reading four hours of documentation before installing anything feels wasteful when you could be running Plex by sundown. The trade is that the exciting choice costs you a weekend the first time something breaks, and a quarter the second time, and the only people who tell you the exciting choice was a mistake are the ones who came out the other side of the rebuild.
The other trade-off is the one between flexibility and simplicity. Proxmox gives you more flexibility than Unraid. Unraid gives you more simplicity than Proxmox. Debian with Docker gives you more transparency than either, at the cost of more glue code you have to write yourself. Pick the trade-off that matches your weekend tolerance, not the trade-off that matches your ambition.
If you take one thing from this article, let it be the boring choice. Install Proxmox or Unraid or Debian with Docker, update it once, walk away from it for an hour. The next session, add one service. Then another. The boring base survives the next three years of tinkering. The exciting base survives the first quarter.