The NVMe in my big HP Workstation died yesterday. I’m not sure why other than it was cheap and used from eBay. I, however, decided to treat this as an opportunity instead of crying in my beard.

Something I’ve planned for a while now, but had been putting off, is a significant refactoring of my home lab. My setup consisted of a huge HP workstation with a Xeon processor and 128GB of ECC RAM. It was entirely overkill for what I used it for, and it was costing some serious money to keep running around the clock.

About a year ago, I bought a pack of Lenovo Thinkcentre m715q mini PCs on eBay. I’ve been using one of the four for my file server, but the other three have been collecting dust. I’ve been thinking about replacing my HP workstation with these little computers for ages, but I never pulled the trigger. There’s something heady about having 128GB of RAM and a beefy processor available to you, even if you don’t use them.

But the death of my SSD prompted me to finally go through the process of downgrading to the smaller, more efficient, mini PCs. I spent most of my day today actually setting this up (when I wasn’t waiting for a furniture delivery at my parent’s house). It was a pain. in. the. ass. But I think it’s going to be worth it.

Here’s what I did.

The Beast⌗

This was my existing file server running on one of the m715qs. I decided that I’d move my Navidrome and Jellyfin setups directly to this computer. I had been running them on the Home lab, and then accessing the files through NFS. It was slow and clunky and heavily network reliant.

With Navidrome and Jellyfin right there with the music/tv/movies libraries, it should make the entire experience faster. (So far, it really has, I’m quite happy)

This will also stay a file server for other things.

Docker1⌗

This is my main server. It houses my Speedtest, uptime kuma, FreshRSS, linkding, and immich containers. I’ll probably add to that list as time goes on.

I chose these, as I trust this server the most. It is in the best shape, has the better NVMe in it, and just feels sturdier.

Docker2⌗

This one I don’t trust as much. It has the NVMe that came with it still inside (I thought I had a few 1GB drives sitting around, but I couldn’t find them), and has some cosmetic damage. I think it will be fine, but I’ll probably end up replacing the SSD at some point.

This one runs my self-built homepage, searx, my Mastodon autoposter, and Booklore. Booklore probably should be on Docker1, but I honestly don’t use it very often, so it isn’t going to hurt if I lose it. We’ll see if this is where it stays.

Docker3⌗

This one isn’t running yet, and may not. I have a HP prebuilt computer with 32GB of RAM and a bunch of storage sitting in my spare room, and I’m thinking about adding this to my cluster.

I’m just not sure what I’d do with it yet, so I’m holding off on it. If my point is to save some electricity, having a full sized PC running along with the other three will defeat the purpose.

VPS⌗

I have a VPS that I’ve been using to host the Matrix/Discord Bridge (thank the Good Lord it wasn’t on my home server, else I’d have lost it and I’d have probably set fire to my desk if that had happened).

Since I’ve paid for it, I’m going to host some things there that I don’t want to lose. Namely, my Castopod instance (though I’m thinking about moving this to Fireside) that I use to host my podcast The 3 Cast, my self-made commenting system for The Linux Cast Blog, my self-made and highly innaccurate website analytics thing, and my cross poster for Mastodon and Bluesky.

This is the only one that I’ve not completed yet. I’m still trying to think about the Castopod thing. I’m lucky that I still have the data for it, so I may just reinstall. But Fireside (where I host The Linux Cast pod) is very good. It would be more expensive, but I’d be able to stop moving from one Castopod instance to another (this is the third time I’ve had to reinstall).

I’ve also lost the data for all my TLC comments, so I need to figure out how to get that back. I have the export from Hyvor so I should be able to get some of it back.

Complicated, But I’m Happy⌗

It definitely is more complicated than it was when it was just the one computer. But together the three of these don’t take up as much power as the HP did, and it’s unlikely that all three will die at the same time, so it should make it a bit more redundant.

I may, over time, replace these with newer mini PCs, but for now they’ll work just fine and I didn’t even have to spend more money on them.

I thought about doing a proxmox cluster, but I just don’t need that type of power. Maybe someday.

What are your thoughts? How do you handle self hosting?