booted old kernel when new one is available

[pbrezina]

It looks like bootc boots the old kernel when upgraded package is already available.

I run dnf upgrade in the Containerfile and new kernel is installed into the image. But bootc upgrade && sudo reboot still loads the old image. It looks like additional image rebuild and bootc upgrade is needed to pick up the updated kernel. Is there any way to start with the latest kernel immediately? Or is this a bug?

[pbrezina /var/home/pbrezina]$ rpm -q kernel
kernel-6.12.7-200.fc41.x86_64
kernel-6.12.8-200.fc41.x86_64
[pbrezina /var/home/pbrezina]$ uname- r
bash: uname-: command not found
[pbrezina /var/home/pbrezina]$ uname -r
6.12.7-200.fc41.x86_64

[cgwalters]

You can't have multiple kernels in a bootc image, only one. I would avoid RUN dnf upgrade in a container build today.

[spmfox]

@cgwalters I've had this question myself - is there any way to install and replace the kernel during the build process? If not currently, is it on the roadmap? Thank you.

[cgwalters]

See coreos/rpm-ostree#5135

[pbrezina]

This is what I use right now:

#!/usr/bin/env bash
#
# Update kernel to latest version and remove old kernels to ensure that
# only the latest kernel is installed.

dnf upgrade -y kernel

KERNELS=`rpm -q --queryformat '%{VERSION}-%{RELEASE}.%{ARCH}\n' kernel`

for kernel in `echo "$KERNELS" | head -n -1`; do
    dnf remove -y kernel\*-$kernel
done

KERNELS=`rpm -q --queryformat '%{VERSION}-%{RELEASE}.%{ARCH} ' kernel`
echo "Installed kernel: $KERNELS"

I'm using fedora bootc on my desktop and I want to stick with running dnf upgrade in the case the base image deployment is delayed for whatever reasons. But it looks like it is currently rebuilt daily, so it's probably not really needed as far as there are no issues with the build process.