CentOS 8, Failed to switch root, '/sysroot' Empty

[Sondergaard]

I just updated my kernel for the first time since my problem. The update succeeded and I can boot without the hack to the file in /boot/loader/entries that I described earlier.

Why the problem has resolved, I don't know for sure. I did run grub2-mkconfig -o /boot/grub2/grub.cfg, but I did that the last time. I only changed one thing, and noticed another.

• What I noticed was that this update included grub2-common-1:2.02-90.el8.noarch. Did this new release of grub2-common make a difference? I don't know.
• What I changed was that I got rid of the symbolic link that ensured /boot/grub2/grubenv and /boot/efi/EFI/centos/grubenv were the same, and instead used identical copies in both places. I was already doing this with /boot/grub2/grub.cfg and /boot/efi/EFI/centos/grub.cfg, as I had found my system failed reboot if I used a symbolic link there. Is this what made the difference? I don't know that either.

Now that I have CentOS 8 working I guess my next job is to replace it, since now it will not be lasting me seven years but only 12 months.

[sml]

What I noticed was that this update included grub2-common-1:2.02-90.el8.noarch. Did this new release of grub2-common make a difference? I don't know.

I suspect this bit in the changelog:

* Di Aug 11 2020 Javier Martinez Canillas <javierm@redhat.com> - 2.02-88
- Fix a legacy BIOS boot issue when using a config file shared with EFI
  Resolves: rhbz#1850193

[TrevorH]

I was already doing this with /boot/grub2/grub.cfg and /boot/efi/EFI/centos/grub.cfg, as I had found my system failed reboot if I used a symbolic link there. Is this what made the difference? I don't know that either.

You should not need to do this. Only one copy of the file is used and it depends on whether your machine is in UEFI mode or not as to which one. If /sys/firmware/efi exists then your machine is in UEFI mode and the file under /boot/efi/EFI/centos/grub.cfg is used, if not then it's /boot/grub2/grub.cfg

[vj82online]

I faced the same issue but was able to fix it after following the post

https://access.redhat.com/solutions/5061041