we are moving forward with the idea of distributing PXE files via OCI registries. We have filed a Fedora Change and we have a working testing pipeline for distributing the files:
The artifacts are extracted and pushed via buildah tool, there is still no download artifact support in podman but upstream is discussing the idea. In the meantime, tool from Fedora named golang-oras can be used to download artifacts:
The namespace is temporary, we will move the repo into pulp project for the time being and hopefully the change is accepted and the final artifacts will be pushed in the official fedora namespace.
We want to do a followup for RHEL and CentOS Stream. Once Fedora Change is accepted and the pipeline is final, including signing with GPG keys, I will update the Foreman repo doing the same thing for non-RH OSes.
So I understand correctly - the big benefit to using these boot/kickstart files as OCI artifacts is that you can avoid having to sync a Kickstart RPM repository? Which then enables the user to perform all automation via native container workflows?
Iām just catching up now on the idea here, but it sounds like Katello should consider this as a core provisioning workflow to support when development shifts to improving the image mode provisioning story.
Just as Katello exposes boot files today from kickstart repositories during provisioning, it could distribute the kickstart artifacts from these new OCI repos.
Yes, the idea is for bootc systems, there are no RPM repositories everything is a container.
Yeah that is the idea, in fact, I just posted a lengthy post that explores HTTPS EFI boot provisioning where PXE files from KS repos are actually made irrelevant (they are unused, only the RPMs) and this is exactly what it needs to effectively download these files into Pulp.