Manjaro quit working today, but it is Friday. Did I deploy my tried and true embedded systems root cause analysis skills? Hell no! It’s Friday and it was working before I turned it off to connect it to a KVM. All I was doing was bringing one other machine back on-line because that client had more work for me. I dutifully powered off the 6-core AMD and cabled everything up through the dual monitor KVM.
Turned this system on. Nothing (&)(*&0ing worked. Well, the fans started, drive light flashed, nothing on the screen though. Swapped cables, slammed head, still not “working.” It had been direct connected to these two monitors before and was working perfectly until I turned it off!
A Tiny Bit of Root Cause Analysis Crept in
Okay, try turning on the Dell the client sent me that has been sitting on a shelf for months. That thing takes forever to boot, but it seemed like nothing was actually working. Fiddled with the connections, one monitor showed the BIOS screen and the initial Windows boot logo before blackness. Progress!
A bit more connection fiddling had that machine displaying on both monitors. It also showed me Windows had helpfully started downloading a zillion GB of Windows updates which was making the machine run like a lawnmower pulling an 8-bottom plow.
You don’t have to know anything about farming to understand that statement now that I’ve shown you an 8-bottom plow. Your favorite badass zero turn ain’t got a prayer of breaking sod with that. That’s how an even great Windows machine runs when you turn it on, connected to the Internet, after months of being turned off. Every (&)(*&)ing package installed on the machine checks for updates and pulls them down.
What do we know?
We know that Rollie shouldn’t be working on shit like this on a Friday. We also know that everything from the KVM to the monitor works. I rebooted many times. Changed cables. I even put in a different video card. The hard drive light would flash a bit, but nothing on either display.
I’ve written about Manjaro on this blog before. It’s quickly becoming my favorite distro. If I didn’t have some development things that needed to use Ubuntu (or at least a YABU) I would have jumped ship to Manjaro as my primary desktop. I had it configured to look really nice on that machine. It was really snappy too, even with BOINC running in background.
The BOINC Rack
I’ve written about the BOINC Rack many times before as well. I have a bench with raised shelf under it where I put my “spare” machines. They run BOINC trying to help make the human species and our world better. It has an older PS/2-VGA single monitor KVM with an older small monitor. What’s important is that that combination works with everything.
Sure enough, I see the “cleaning” message Linux likes to give you when it wasn’t shut down just proper. Hard drive light goes spinning, after a few minutes goes quiet, then starts to intermittently blink.
Did I say to myself “Gee, that’s what it looks like when BOINC is running?” No! Did I think “Gee, I wonder if I can <Alt><Ctrl><F3> and log in via terminal?” No! I got out a GParted CD and checked the partition. Checked other things with SystemRescueCD. I even formatted the thing and restored from last weekend’s backup. Same Shit Different Day.
Your Kernel Killed You
That system had a GT218  NVidia card. Actually both of the ones I was trying. I forget which third party made this one. The kernel finally got new enough that there is no longer a driver from NVidia to support the card. Did the kernel update back out the NVidia driver and leave me with a usable though CUDA-free system? No! It left the non-functioning driver in place and the video driver gets loaded early in the boot process so all I could ever see was the “cleaning” message.
Never Buy an AOC Monitor
Why did I see nothing with the other KVM and monitors? Well, one of those monitors is an AOC. Never buy an AOC. Never use an AOC with NVidia anything. I’ve never figured out why, but during boot, every NVidia card I have ever had seeks out the AOC to use as its primary for boot. Doesn’t matter what connector I hook it to. That and probably every other AOC monitor ever made has a “unique power saving feature:” they don’t display BIOS messages. They seem to wait for the video driver to get loaded and swat them to life.
Manjaro quit working because the new kernel took effect at boot. This kernel was incompatible with the NVidia driver and didn’t bother to sweep that up.
I have Mint LMDE 5 running on that box now and I bought a used Quadro 2000 (uses the still supported 390 driver) on eBay for just over $20. Manjaro may yet find its way back on the box because it is a much faster Linux distro. All of the Debian based stuff seems to run like half-set concrete pours these days.