Describe the bug
Title
SCOS 4.22 upgrade leaves system booting to grub> prompt because EFI/centos/grub.cfg is missing
Body
Describe the bug
After upgrading an OKD worker to 4.22.0-okd-scos.5, the node no longer booted automatically.
Instead, the machine stopped at a bare GRUB prompt:
The OS deployment itself was healthy and could be booted manually.
The problem was eventually traced to the EFI partition: the firmware booted \EFI\centos\shimx64.efi, but there was no accompanying \EFI\centos\grub.cfg (nor bootuuid.cfg).
Creating these files restored normal unattended boot.
Environment
CentOS Stream CoreOS 10.0.20260614-0
Kernel 6.12.0-233.el10.x86_64
Symptoms
After reboot:
No menu appeared.
The system could be booted manually using:
set root=(hd1,gpt3)
set prefix=(hd1,gpt3)/grub2
insmod normal
normal
After that, GRUB loaded normally and booted the installed SCOS deployment.
Investigation
The deployed OSTree deployment was healthy.
showed:
State: idle
Version: 10.0.20260614-0
ostree admin status:
bootupctl status:
Component EFI
Installed: grub2-efi-x64
Update: At latest version
The BLS configuration was present:
contains:
The BLS entry:
/boot/loader.1/entries/ostree-1.conf
contained the expected kernel and ostree arguments.
EFI contents
EFI partition contained:
EFI/
├── BOOT/
├── centos/
│ grubx64.efi
│ shimx64.efi
│ shim.efi
│ ...
└── fedora/
grub.cfg
bootuuid.cfg
Notably missing:
EFI/centos/grub.cfg
EFI/centos/bootuuid.cfg
Workaround
Copying the Fedora stub configuration into the CentOS directory fixed the problem:
mount /dev/<efi-partition> /mnt/efi
cp /mnt/efi/EFI/fedora/grub.cfg \
/mnt/efi/EFI/centos/grub.cfg
cp /mnt/efi/EFI/fedora/bootuuid.cfg \
/mnt/efi/EFI/centos/bootuuid.cfg
After reboot the machine booted normally without manual GRUB interaction.
Current status after workaround
BootCurrent: 0004
BootOrder: 0004,...
Kernel command line:
ostree=/ostree/boot.1/fedora-coreos/...
Node boots normally.
Question
Is the SCOS migration expected to populate:
EFI/centos/grub.cfg
EFI/centos/bootuuid.cfg
when the firmware boot entry points to:
or is this a bug in the migration/update process?
I'd also attach:
efibootmgr -v
ls -R /boot
ls -R /mnt/efi/EFI
rpm-ostree status
bootupctl status
Those will give the maintainers everything they need to reproduce or diagnose the issue.
One final note: because this occurred immediately after upgrading to 4.22.0-okd-scos.5, it would also be worth checking whether other users have reported similar SCOS boot issues before filing a new issue, and if not, this report should provide enough detail to start the investigation.
Version
4.22.0-okd-scos.5
Reproducibility
No response
Log Bundle
No response
Describe the bug
Title
SCOS 4.22 upgrade leaves system booting to
grub>prompt becauseEFI/centos/grub.cfgis missingBody
Describe the bug
After upgrading an OKD worker to 4.22.0-okd-scos.5, the node no longer booted automatically.
Instead, the machine stopped at a bare GRUB prompt:
The OS deployment itself was healthy and could be booted manually.
The problem was eventually traced to the EFI partition: the firmware booted
\EFI\centos\shimx64.efi, but there was no accompanying\EFI\centos\grub.cfg(norbootuuid.cfg).Creating these files restored normal unattended boot.
Environment
OKD release:
4.22.0-okd-scos.5SCOS image:
Symptoms
After reboot:
No menu appeared.
The system could be booted manually using:
After that, GRUB loaded normally and booted the installed SCOS deployment.
Investigation
The deployed OSTree deployment was healthy.
showed:
ostree admin status:bootupctl status:The BLS configuration was present:
contains:
The BLS entry:
contained the expected kernel and ostree arguments.
EFI contents
EFI partition contained:
Notably missing:
Workaround
Copying the Fedora stub configuration into the CentOS directory fixed the problem:
After reboot the machine booted normally without manual GRUB interaction.
Current status after workaround
Kernel command line:
Node boots normally.
Question
Is the SCOS migration expected to populate:
when the firmware boot entry points to:
or is this a bug in the migration/update process?
I'd also attach:
efibootmgr -vls -R /bootls -R /mnt/efi/EFIrpm-ostree statusbootupctl statusThose will give the maintainers everything they need to reproduce or diagnose the issue.
One final note: because this occurred immediately after upgrading to 4.22.0-okd-scos.5, it would also be worth checking whether other users have reported similar SCOS boot issues before filing a new issue, and if not, this report should provide enough detail to start the investigation.
Version
4.22.0-okd-scos.5
Reproducibility
No response
Log Bundle
No response