The title of the post has it all. The first steps to migrate from CentOS 7 to Rocky Linux 8 fail (in my eyes) because of module RPMs.
If something is installed the traditional way via dnf, dnf notices if a dependency is only available as a modular RPM and activates the the app stream of the module to install other dependencies in the correct version.
For automatic installation of a customized distribution, additional RPMs are downloaded via --download-only of dnf. In this specific case, docker is installed on Rocky Linux 8. Module RPMs and dependencies are decoupled from their stream and put into the ISO so that the automatic installation works without problems.
If a module RPM is found, repo2module is called to create a module.yaml. The YAML file is dropped into the additional RPM repo in the ISO image and the repository is updated via modifyrepo_c. Otherwise the module RPM repository will not work.
If the distribution was installed successfully it is not recognized that some installed module RPMs belong to an existing app stream. The were previously decoupled via --download-only.
If you try to activate the app stream of the RPM module manually, dnf detects dependency problems with the already installed RPMs.
Now a system was installed, which cannot be updated increasingly. For this reason, the offline installation of a customized RHEL based distribution via kickstart is currently not possible in my eyes, because updating the system is not possible with an existing connection to an RPM server.
- How did the developers envision the handling of kickstart distributions with module RPMs? Maybe I make a mistake in the compilation of the distribution.
- How to fix the problem with updating pre-installed module RPMs?
