Description of problem: Client on OSP13 with UEFI nodes used grubby on its compute nodes to modify an option. Client did a FFU 13 to 16.1 last fall. Yesterday they did a minor update from 16.1.5 to 16.2.1. On reboot of those compute nodes they would boot the RHEL8.2 kernel even though 'grubby --default-kernel' would show 8.4 being the default one. To fix this we did the following: rm /boot/efi/EFI/redhat/grubenv mv /boot/grub2/grubenv /boot/efi/EFI/redhat/grubenv ln -s /boot/grub2/grubenv /boot/efi/EFI/redhat/grubenv And reboot would now default to RHEL 8.4 kernel. A patach has been written here and might be fixing this situation: https://review.opendev.org/c/openstack/tripleo-heat-templates/+/826483 We have sosreport of compute nodes having this issue. Version-Release number of selected component (if applicable): OSP16.2.1 How reproducible: 100% Steps to Reproduce: 1. OSP13 + grubby to modify the config 2. FFU to 16.1 3. Minor to 16.2 4. Reboot node. Actual results: Node would boot RHEL 8.2 kernel instead of new RHEL 8.4 kernel. Expected results: Boot the new RHEL 8.4 kernel. Additional info: sosreport are attached to another case. Will put that information in next private comment.
I ended up circling back to this due to 16->17 upgrade. The copy of grubenv instead of linking was due to some grub bug which probably all envs no longer hit so we should patch update and upgrade in 16.2.