This project concluded months ago, and this is more a postmortem than a formal writeup. Future projects will be written up more properly.

The one thing I hate more than anything else is being bored.

I’ve been using Docker Compose for years at this point, and while it was initially painful to learn, it quickly became the most natural way for me to compose and view my services at a glance. Docker is truly a wonderful piece of software, and I love to use it. Eventually though, as happens with all my projects, I eventually got a bit too comfortable using it. That is something I couldn’t stand for.

For some time, I had been hearing about this thing called Kubernetes. I talked to my friends about it, they said they hated it. I asked my brother-in-law (fellow homelabber), he said it was “cool, but when would you use it?”. Nobody’s negative opinions made me want to try it less— if anything, I was more interested now than ever.

The sane route for this would almost certainly be to spin up several VMs on my workstation, wire them up, play around. However, I rarely pick the sane option (for anything, ever), and after crawling eBay, I found a sweet deal on some HP t530 thin clients. No drives, no RAM, but cheap! 4 for like, $50. I scoop them up as fast as I can, not even considering why they might be so cheap. The goal was to get them working, wire them up with k8s, run my services off of them, and be on my merry way.

I could not have been more wrong about how convoluted and stupid this entire project would end up being.

My thin clients arrive, I install RAM and SSDs into them, boot them up— no display. I try different DisplayPort adapters, no dice. After some further poking around (and the blinking ethernet port + constant heating up of the machine), I came to the conclusion it had been locked into some sort of PXE-only boot.

This led me down another rabbit hole where I learned how PXE worked, retooled my homelab’s networking stack (a completely different rabbit hole involving an HP Aruba switch I bought and had to learn to operate) so that I could fit a PXE server into the setup at a low enough level that even the absolute dumbest of machines would actually receive the signal. No dice, still.

It’s at this point in my sisyphean journey that I start becoming increasingly irate about the entire affair. I decide to do some network sleuthing and see just what on God’s green Earth these thin clients could possibly be doing that is making them heat up so much when they won’t even do anything other than power on and blink some pretty lights. I find something important: they are just screaming for HPDM (HP Device Manager, HP’s proprietary IT administration software). At this point, I’m so happy to have any solution in sight that I don’t even kick myself for assuming incorrectly that PXE was what was needed. I go through all the hoops to download HPDM, dust off my Windows dual boot on one of my computers, install it, fire it up, and…

…nothing. Nothing! Nothing, at all. The extra thin client of the same make and model that I purchased in fully working condition is detected just fine, but the others? The ones that started this entire project? None.

So we’ve now determined what signal they are looking for (HPDM), but they clearly are not looking for a standard one. So what are they looking for? To answer this, a sidequest I glossed over:

At some point during the project, I purchased a fifth thin client. A working one, shown to be hooked up to a display with Windows 10 pre-installed. I got it, plugged it in, it just worked! I had an idea: I hooked up my BIOS dumper/flasher, dumped the BIOS off the good unit, flashed it onto a bad unit and: no reaction! I dump the BIOS off a second thin client, and run a diff on the two files. They are the same. They are the same file. The BIOS should, in theory, work. Critically though, it does not.

I have nothing left in this project except for more hypotheses that I am not personally skilled enough to check. Between the price point of the thin clients and their strange behavior, my main thesis is this: these thin clients were never sold, never prepped for sale, and as such are not ‘activated’. Many HP devices have something in their pre-boot environment that lists the device’s serial number, among other things, which then authorizes the device to boot. These devices most likely do not have that data. They don’t even have serial numbers listed on the devices themselves.

I have a lot of questions about how these ended up on eBay in the first place. Are they stolen? Were they ‘thrown away’, and grabbed by an employee?

I really do not know. E-mail me if you have an idea of how to dump and flash the contents NVRAM/EEPROM off of a HP-T530 thin client, otherwise I’ll see if I can’t find anyone more knowledgeable on this subject at some point in time for guidance. I am not done with this necromancy project yet, mark my words.