The overlay was active when the system was suspended.
The system has auto-updates enabled, and had not been rebooted in several days, so it’s likely that there was an update queued for the next reboot.
2. Learn how to recover the system from its current state
So far, I’ve tried this from a live USB:
sudo -s
cryptsetup luksOpen /dev/sda3 encrypted-partition
mount -o subvol=root /dev/mapper/encrypted-partition /mnt
cd /mnt/ostree/deploy/fedora/deploy/xxxx.0 # There are multiple deployments. I just picked the first one
mount /dev/sda2 boot
mount /dev/sda1 boot/efi
mount --bind /dev dev
mount --bind /dev/pts dev/pts
mount --bind /proc proc
mount --bind /sys sys
mount -o subvol=var /dev/mapper/encrypted-partition var
mount -o subvol=root /dev/mapper/encrypted-partition sysroot
chroot .
grub2-mkconfig -o /etc/grub2-efi.cfg
# /usr/sbin/grub2-probe: error: cannot find a device for / (is /dev mounted?).
Hello @tbrown ,
Do you get the boot menu? If not enter your BIOS at boot or use your BIOS’s boot menu to select the fedora boot option. If you can boot into your existing system then you could use rpm-ostree rollback to boot your previous commit and upgrade it. The default commit is #0 from Ostree’s POV.
Should never be used on Silverblue. The boot entries are only modified by rpm-ostree.
While troubleshooting I think I might have modified the boot and efi partitions, so I decided to follow your advice here and reinstall Silverblue while keeping my home subvolume.
This worked perfectly, and I was back with my original working system! (minus a few overlaid packages)
The UEFI boot menu then contained two entries:
AHCI P2: Samsung SSD 870 QVO 1T
Fedora
But then I wondered if I would be able to reproduce the behavior.
I booted into the Fedora entry, suspended the system, and cut the power - and now it’s again in the same unbootable state! This was on a freshly installed Silverblue 39.1.5 with no overlays or pending updates showing in rpm-ostree status. Although I used the advanced custom partitioning option during install (to keep my home btrfs subvolume), I tried to replicate the original partitioning and mounts.
Now the UEFI boot menu only contains one entry:
AHCI P2: Samsung SSD 870 QVO 1T
and when I select it I get this message:
Reboot and Select proper Boot device
or Insert Boot Media in selected Boot device and press a key
No GRUB boot menu in sight.
I’m curious what is causing it to break, but I have no idea how to debug it. If anyone wants any more information please let me know.
Otherwise I guess I’ll just fix it up and hope my power doesn’t go out again!
Well that’s really weird isn’t it? Suspend —> kill power —> Lose boot. Can you please file a bug at Issues · fedora-silverblue/issue-tracker · GitHub so we can get the Silverblue team helping? Link the issue back here for reference and we’ll keep this topic going till resolution. Sound good?
Turns out the CMOS battery on my motherboard was flat, which was causing the Fedora entry in the UEFI boot menu to get forgotten when the motherboard lost power