Blivet GUI which was the ancestor of cockpit-storage in the anaconda context always created partitions on top of disks, when the user tried to create a Software RAID device. Therefore Software RAID on top of disks was never tested or supported in anaconda. Currently this causes unbootable systems, because RAID created from cockpit have metadata version 1.2 but efi bootloader has limitation to 0.9 and 1.0 metadata versions. See https://bugzilla.redhat.com/show_bug.cgi?id=2352953 for more info. Let's disable this feature (Creating MDRAID on top of disks directly) for Fedora 42 for anaconda mode, and possibly have time to properly test it and support this for fedora 43. Reproducible: Always
(In reply to Katerina Koukiou from comment #0) > Currently this causes unbootable systems, because RAID created from cockpit > have metadata version 1.2 but efi bootloader has limitation to 0.9 and 1.0 > metadata versions. What happens when I create a MDRAID with metadata version 1.2 on partitions? Will that boot, or will Anaconda then correctly detect that it is invalid and refuse to proceed? > See https://bugzilla.redhat.com/show_bug.cgi?id=2352953 for more info. That talks about BIOS and biosboot, not EFI. Does bios also have problems with the MDRAID 1.2 format? > Let's disable this feature (Creating MDRAID on top of disks directly) for > Fedora 42 for anaconda mode, and possibly have time to properly test it and > support this for fedora 43. Cockpit already has a "block device preparation" step when creating MDRAIDs. Currently this is used to create a partition for the free space in a existing partition table. We should be able to extend this pretty easily to also create a partition table plus partition without major code surgery.
Bug 2352953 is very similar (I'd even say same root cause), so I think this can be closed as a duplicate?