My way of dealing with this is to ensure everything is provisioned and managed via gitops. I have a homelab repo with a combination of Ansible, Terraform (Tofu), and FluxCD. I don't have to remember how to do anything manually, except for provisioning a new bare metal machine (I have a readme file and a couple of scripts for that).

I accidentally gave myself the opportunity to test out my automations when I decided I wanted to rename my k8s nodes (FQDN rather than just hostname). When I did that, everything broke, and I decided it would be easier to simply re-provision than to troubleshoot. I was up and running with completely rebuilt nodes in around an hour.

But configuring a FreeBSD system with zfs and samba is dead easy.

In my experience, a vanilla install and some daemons sprinkled on top works better than these GUI flavours.

Less breakage, fewer quirks, more secure.

YMMV and I’m not saying you’re wrong - just my experience

Yeah; once you get deep enough, you realize you can just install things yourself, configure them with Ansible, or Nix, or whatever, and have full control.

But for probably 90% of users, they just want a UI they can click through and mostly use the defaults, then log into now and then to make sure things are good.

The UI is also especially helpful for focusing on things that matter, like setting up scrubs, notifications, etc. (though even there I think TrueNAS could do better).

It's why Synology persists, despite growing more and more hostile to their NAS owners.

> I can setup Samba, NFS, and ZFS manually myself, but why would I want to? Configuring samba users & shares via SSH sucks. It's tedious. It's error prone. It's boring.

I agree for the most part, though even vanilla Ubuntu has Cockpit if you need a GUI.

Personally I find that getting it set up with NixOS is pretty straightforward and it's "set and forget", and generally it's not too hard to find configurations you can just copypaste done "correctly". And of course, if you break something you can just reboot and choose a previous generation. Of course restarting still requires SSHing in and `systemctl restart myapp`, so YMMV.

I want to configure it myself because now I know exactly how it works. The configuration options I’ve chosen won’t change unless I change them. Disaster recovery will be easy because when I move the disks to a new machine, LVM will just start working.

I actually can’t recall the last time I setup a share on SMB, has to be years if not tending towards decades.

A few big shares is all I really need; I no longer create a share for every single idea/thing I can think of.

I’ll take Configuration Management for $100, Alex.

Even though I have over four hundred Blu-rays, I would of course NEVER condone breaking the DRM and putting them on Jellyfin no matter how easy it is or how stupid I think that law is because that would be a crime according to the DMCA and I'm a good boy who would never ever break the law.

That said, I have lots of home movies that just so happen to be at the exact same bitrates as Blu-rays and after the initial setup, I've never really had any issues with them choking or any bandwidth weirdness. Minecraft doesn't use a ton of disk IO, especially since it is rare that anyone plays on my server other than me.

I do occasionally do stuff that requires decent bandwidth though, enough to saturate a WiFi connection at the very least, and the USB3 + USB SS + Thunderbolt never seems to have much of an issue getting to Wifi speeds.

Nope!

I have one enclosure plugged into a USB 3.0 line, another plugged into a "super speed" line, and one plugged into a Thunderbolt line (shared with my 10GbE Thunderbolt card with a 2x20 gigabit splitter).

This was deliberate, each is actually on a separate USB controller. Assuming I'm bottlenecked by the slowest, I'm limited to 5 gigabits per RAID, but for spinners that's really not that bad.

ETA: It's just a soft RAID with ZFS, I set up each 8-bay enclosure with its own RAIDZ2, and then glued all three together into one big pool mounted at `/tank`. I had to do a bit of systemd chicanery with NixOS to make sure it mounted after the USB stuff started, but that was like ten lines of config I do exactly once, so it wasn't a big deal.

> How do you fill 288 TB? Is it mostly media?

I kind of purposefully don't fill it up :).

This has been a project of tech hoarding over the last ~8 years, but basically I wanted "infinite storage". I wanted to be able to do pretty much any project and know that no matter how crazy I am, I'll have enough space for it. Thus far, even with all my media and AI models and stock data and whatnot, I'm sitting around ~45TB.

On the off chance that I do start running low on space, there's plenty of stuff I can delete if I need to, but of course I probably won't need to for the foreseeable future.

> I'm converting more of my infrastructure to NixOS, and I know a lot of people just manage their NAS with NixOS, which is sounding more and more appealing to me.

Yeah, that's what I do, I have my NixOS run a Samba on my RAID, and it works fine. It was like fifteen lines of config, version controlled, and I haven't thought about it in months.

288TB might be enough to store a complete set of laser disk arcade games. /s

I think that's a good idea. It's a more manual setup but also more space efficient (if you skip backing up eg linux isos) and causes less wear than RAID.

> Do you think the power consumption matters on your box here?

It's actually not too bad; the main "server" idles at around 14W and the power supply for it only goes to 100W under load. The drive bays go up to 100W (I think) but generally idle around 20W each. All together it idles at around ~70-80W.

Not that impressive BUT it replaced a big rack mount server that idled at about 250W and would go up to a kilowatt under load.

> Should you care about the "USB bottleneck"?

Not really, at least not for what I'm doing. I generally can get pretty decent speeds and I think network is often the bottleneck more than the drives themselves.

> How do you organize this thing so it's not a mess of USB cables?

I don't :). It's a big mess of USB cables that's hidden in a closet. It doesn't look pretty at all.

You... swap the drive, run a ZFS replace command, and that's it. I know I'm coming at this from a particular perspective, but what am I missing?

I tried a Mac Mini for a while, but they're just not designed to run headless and I ultimately abandoned it because I wanted it to either work or be fixable remotely. Issues I had:

- External enclosure disconnected frequently (this is more of an issue with the enclosure and its chipset, but I bought a reputable one)

- Many services can't start without being logged in

- If you want to use FileVault, you'll have to input your password when you reboot

Little things went wrong too frequently that needed an attended fix.

If you go off the shelf, I recommend Synology, but make sure you get an Intel model with QSV if you plan to transcode video. You can also install Synology OS to your own hardware using Xpenology - its surprisingly stable, moreso than the mac mini was for me.

I do recommend those little Beelink computers with an AMD CPU.

They can be had for a bit less than the Mac mini, and I had no issues getting headless Linux working on there. I even have hardware transcoding in Jellyfin working with VAAPI. I think it cost me about $400.

I was in the same boat - QNAP OS was a complete mess. Ended up nuking it and throwing Ubuntu on there instead. Nothing fancy, just basic config, but it actually works now. Other option is pay Unraid.

What are your requirements for a NAS? What do you want to use it for? Is it just to experiment with what a NAS is, or do you have a specific need like multiple computers needing common, fast storage?

I use a QNAP 8-bay JBOD enclosure (TL-D800S) connected via SFF-8088 to a mini-itx PC - I find the form factor pretty good and don't have to deal with QNAP OS.

The fact that Asahi doesn’t run on M4 cpus (ie the current Mac Mini) is also a consideration.

ZFS on macOS sucks really bad, too, so that rules out the obvious alternative.

I've had mixed luck too. The one I'm using now has been mostly ok. https://a.co/d/4AiF1Zp

It actually has been considerably more reliable than the MediaSonics than they replaced.

Ive seen a lot of people linking beelink mini PCs lately: https://www.bee-link.com/products/beelink-me-mini-n150

The cost is very low, but it would have to be a nvme only build.

On a somewhat related note, would you trust a Pi based NAS long term? I've not tried doing one since the Pi 4 which understandably because of its hardware limitations left a lot to be desired, but that part aside I was still finding the pi as a piece of hardware somewhat quirky and unpredictable - power especially, I can't count the number of times simply unplugging a usb keyboard would cause it to reboot.

totally agree. it becomes extremely obvious when applications are poorly optimized or have outdated build systems that don't support ARM.

imo, if the software doesn't work without issue on my Pi, it isn't good enough for prod.

I am curious about the slow part. I use these crappy SATA cards and I am sure they are crappy, but the drives are only going to give 100MB/s in bursts and they have an LVM cache (or ZFS stuff) on them to sustain more short-term writes.

I get if I was wiring up NVME drives that are going to go 500MB/s and higher all the time.

What I really care about with the SATA and what I mean by flaky is I when I have to reboot a system physically every day because the controller stays on in some way even if it gets a soft `reboot` command and then Linux fills up with IO timeouts because the controller seems to stop working after X amount of time.

> If your build allows the extra money for an LSI or real raid controller is well worth it.

Keep in mind if you get a real RAID controller and want to use ZFS, you probably want to ensure it can present the disks to the OS as JBOD. Sometimes this requires flashing a non-RAID firmware.

That said, slightly older LSI cards are quite cheap on eBay, and the two I've bought have worked perfectly for many years.

Thank you for the correction.

i have had a random $10 sata card, has worked fine over the last 5 yrs

I do too. For many use cases it's awesome to use an ESP32, Raspberry Pi or Arduino when I want to stash some little widget that can sip a small battery over the next week. It's equally awesome that in many scenarios you can be net positive in your consumption with a super simple solar panel that provides a few watts here and there into a battery.

But at home things are different. While I want to use as little power as possible, the realistic plan for being sustainable at home is to use solar with batteries. That's a plan I actually think can matter and that I am able to participate in for relatively low cost ($10k)

Messing around with a system to save a few watts for me in this context isn't very valuable.

Yeah this is what keeps me from considering old PCs for NAS.

Maybe stating the blindingly obvious but seems like there is a gap in the market for a board or full kit with a high efficiency ~1-10W CPU and a bunch of SATA and PCIe ports.

Probably not very good. I selected large spinning hard drives because I could get them at a good price for 2TB each and I wanted to setup a RAID5-like system in ZFS and btrfs (lesson learned, btrfs doesn't actually support this correctly) and I wanted to get at least 10TB with redundancy.

I don't know how much each of those SATA disks take up, but probably more than a single Raspberry Pi does.

Likewise it has a few case fans in it that may be pointless. I would prefer it never has a heating issue versus saving a few "whurrr" sounds off in a closet somewhere that nobody cares about.

It's also powering that Nvidia 1060 that I do almost nothing with on the NAS. I don't even bother to enable the Jellyfin GPU transcoding configuration because I prefer to keep my library encoded in h264 right now anyhow as I have not yet made the leap to a newer codec because the different smart TVs have varying support. And sometimes my daughter has a friend come over that has a weird Amazon tablet thing that only does a subset of things correctly.

The 1060 isn't an amazing card really, but it could do some basic Ollama inference if I wanted. I think it has 6GB of memory, which is pretty low, but usable for some small LLMs.

Wasn't there a post somewhere on HN yesterday about how slowing down your programs can help you catch problems? Using low end hardware is an automatic way of forcing that :)

Aoostar, Ugreen and Minisforum make mini PC style NAS boxes with 3.5" hard drive bays. I think it's the closest to what you're describing.

The mini itx boards are also interesting, but more if you're actually building the machine yourself. I'm typing this from a 16 core AMD one.

Yeah, that’s why I was asking. Most of the mini-PC market is either doing “the cheapest possible desktop” or “the cheapest possible gaming box”.

Well, I am curious about compact SATA options, and have a peer response to yours that is eminently useful, so… I’d say it’s still on topic.

I don't know how you got that idea when I explicitly say it was a hardware issue.

Also the Odroid-H4+ supports IBECC, that means, ECC parity check done by the CPU using normal non-ECC RAM modules. Very suitable for ZFS/TrueNAS.

Any good cases for that? I’d be afraid of ending up with a lump of duct-taped SATA SSDs around the PCB…

Thanks! I’ve used ODROID devices in the past, but wasn’t aware of that variant.

I do wish they'd spec-bump to Twin Lake CPUs soon.

Indeed, it should be doable.

Cockpit hasn't really improved in a while, though, and although I greatly appreciate 45Drives' committment to it, last time I tried to install their stuff I had a lot of issues with deprecated dependencies...

So I just went "raw" smbd and never looked back, but then again I've been running Samba for almost two decades now and configuring it is almost second nature to me (I started doing it as an alternative to Novell/IPX, and later to replace AppleTalk...)

In practice, I've found that worked well because I very seldom had to do any changes to a file server once it was set up (adding shares was mostly a matter of cloning configs, and what minimal user setup there needed to be done in the enterprise was deferred to SAMBA/CIFS glue).

I would imagine it's because it makes for a lot more fun support possibilities when all the underlying stuff in the stack (kernel, ZFS, Docker, Python, etc. etc.) is subject to the whims of the end user. When you ship the entire OS yourself you can be more certain about the versions of all the kernel+userland stuff and therefore the interactions between them all.

I'm not convinced that TrueNAS Scale is an improvement. I won't make the case that stability and maturity hamper Scale as a whole, but there are definitely discrepancies for SMB performance and limitations in and the use of Kubernetes really overcomplicates things for home users (not saying that TrueNAS Scale Apps were great). I had a version upgrade of Scale catastrophically fail on one system because the Kubernetes implementation couldn't be reconciled during the upgrade (I had no enterprise support, community support had no input, so I was forced to restart from scratch).

You also need a dedicated OS Drive for TrueNAS, which is reasonable in principle for critical systems, but doesn't always really meet the needs of home users with limited drive bays.

Sometimes certain product lines act like they consider customers on rare occasion...

But most manufacturers try to lock the firmware down, and users only get a small subset of configuration menus. For example, the Gigabyte rtx based laptops require patching a machine specific bios to even gain access to the oem firmware areas.

Mostly the modern builds just created a bunch a problems nobody wanted, and didn't improve anything as Asus, Gigabyte, and Razer recently showed.

https://www.youtube.com/watch?v=4er6kD-pxZs

https://www.youtube.com/watch?v=KhfqhCxqpQ8

If you are running signed code on many machines. YMMV... Raspberry Pi avoided the signed code features built into most Broadcom ARM chips for good reasons. =3

Tbf I wasn’t clear, I meant the computer as a network box. When we use the tv and want to watch a stored video, as long as the computer is on the media is accessible.