New Server Part 1

It all started, really, with two bad habits meeting at the worst possible time: tight cash flow and terrible digital hygiene. I don’t back up the way I should, and I don’t have that automatic reflex of, “Hey, maybe back this container up before smashing the update button.” Now I’m sitting here with my main Proxmox node on life support, bootable recovery USB in hand, trying to coax it back to life.

The situation feels like a pretzel — and I’m the one twisted up.

KaraKeep was the first domino to wobble. A database schema migration failed, and suddenly I found myself consulting half-remembered SQL concepts like I’m a surgeon who’s only seen a few YouTube tutorials. Then Tududi followed, database corrupted after an update. That one I’m just letting go — Jotty is its spiritual successor now.

In the middle of this chaos I decided, now seems like a great time to learn about clusters. As a requirement a new member to a cluster needs to be empty. So, I tried convincing a new member/node that it had no containers at all, only to realize again the extra effort to backup, destroy, and revive containers would have worked best.

And, of course, the backup solution that would have saved me — Proxmox Backup Server — is still sitting in the “pending” pile, waiting for its turn on the Gen9 DL380. The irony isn’t lost on me.

Meanwhile, the services I built for myself — the ones meant to free me from commercial apps — are quiet, and I realize just how much I’d come to rely on them. It turns out “flying by the seat of my pants” is thrilling right up until the plane starts losing altitude.

Then temptation arrived.

An eBay listing, glowing like a neon sign in a rainy alleyway: Xeon E-2176G, 6 cores 12 threads, 32 GB DDR4 ECC, ASUS P11C-M/4L — all for $350. Part of my brain said, that’s suspiciously good. The other part said, hush, we’re shopping. I could even reuse my Quadro P2200 if the case plays nice — not counting on it, but it’s fun to imagine.

It would be a serious upgrade from my current “rescued from the bin” desktop: AMD A6-7400K with 24 GB of DDR3. That poor CPU wheezes if I so much as whisper “transcoding” near it. I even managed to pass the GPU through to LXCs, but FileFlows looked at it and said, “No thanks, I’ll just burn the CPU alive instead.”

Migration plans spin around my head like loose screws: straight PVE migration with a new PSU… or raw data backups and a slow, careful rebuild. At least I finally invested in a pair of NVMe drives for OS redundancy. This drive failure felt like the kind of lesson the universe repeats until you actually learn it.

The decline was gradual and noisy. First the network started dropping in early December — headless server, no easy video output, me stubbornly refusing to swap cables. A forced shutdown brought it back, for a while. Then again. A few resets later, kernel panics arrived, some read only complaints, and whispers that Docker might be flapping the network. Each time, another forced reset was like kicking a vending machine harder because it didn’t listen the first time.

Eventually even fsck felt like CPR on a failing heart. The final stage was I/O errors everywhere and the kernel effectively screaming that it couldn’t write to the drive anymore.

And that’s where I am: standing between rebuilding, learning the hard lessons about backups, and clicking “Buy Now” on hardware that feels like both a lifeline and a temptation.

  • Disclaimer: I used ChatGPT to clean up my thoughts in a story format.

  • PS: I used Christmas money to fund most of this project. So, thank you grandparents and parents!

———

I am excited to start building system today!