Weird install behavior with multiple potential target drives

I think there is something in the 4.2.0rc5 install process that steps on drives that are not targeted in the install procedure. I’m not sure precisely how to reproduce this but I’ve seen various iterations of it half a dozen times this weekend.

I have an HP Z420 that runs Qubes just fine. It will boot from an internal SATA SSD, a SATA SSD in a USB enclosure, and an LSI PCIe flash drive.

I’ve done a bunch of installs, messing with LVM and BTRFS, finally settling on BTRFS as the right way to do things. I had a working install on the LSI drive and I left it in place, booting with a thumb drive, and installing Qubes on the USB SATA carrier. I was thinking to make the USB boot device for my main machine and see how Qubes acts with an Nvidia GPU and a 4k display.

I downed the machine, removed the USB stuff, restarted it, and I’m looking at a black screen with a single underscore cursor flashing at the upper left hand corner. I let it sit for several minutes, whatever that was it’s fatal, so the install is running again.

The LSI card with BTRFS in the #2 machine is the way things are going to be for the foreseeable future, can’t risk stepping on my production machine. I have some other things I want to do - VPN, USB tethered phones, making sure I can backup/restore stuff on the machine.

One thing I’ve noticed is that, in addition to the mysterious boot failures sans error messages, there’s nothing like a rescue option when booting the installer. Is there a body of information on troubleshooting stuff like this? If I were going to have a group of people out in the field using Qubes there would need to be something like this. I’m sure people will cringe at the thought, but a live CD OS with the right tools to fix things and an install of something like Anydesk would be great - a sort of poor man’s iDRAC for doctoring downed systems from afar.

Did you try switching to your console (CTRL + ALT + F2)? In my experience a blinking underscore just means the graphical login screen failed to load (lightdm.service). You should have been able to login anyways and check your journalctl to see what went wrong.

1 Like

I have NOT done that and I’m a trifle embarrassed now that you mention it. Thanks for the tip :wink: