When we want to make GPT the default partitioning, we must ensure the installability of a software raid as well. At the same time, we must continue to maintain the availability of a software raid even with MBR (DOS) partitioning.
Testcase 1: GPT
Expected result 1. Anaconda should install a biosboot partition on all disks included in the boot raid. 2. The system should continue to boot with any one of the disks broken (removed), specifically w/o the original start disk (sda). 3. The system will remain bootable even if any single of the RAID member disks is defective.
Testcase 2: MBR
Expected result 1. Anaconda should install a MBR record on all disks included in the boot raid (check log file). 2. The system should continue to boot with any one of the disks broken (removed), specifically w/o the original start disk (sda). 3. The system will remain bootable even if any single of the RAID member disks is defective.
Addendum The actual test results are described in a comment further down! (Comment 6)
Metadata Update from @pboy: - Issue tagged with: in progress
Issue tagged with: in progress
Just to note, according to @chrismurphy, this doesn't work at all right now. In MBR mode, we're not installing MBRs on each disk and setting up the boot partition correctly, etc. In GPT mode, we're not installing BIOS boot partitions on each disk, etc.
According to my findings with F35 or F34 (dont't remember exactly) Anaconda installs the mbr on all disks where /boot is part of a raid partition. You find the corresponding entries in the log and the installation also completes without error messages (unlike in case of GPT).
Metadata Update from @pboy: - Issue tagged with: meeting
I've copied this bug to https://bugzilla.redhat.com/show_bug.cgi?id=2092116 and proposed it as a blocker for F37 Final.
Note that you cannot have MBR on BIOS systems in some circumstances already (when the drive is >2TB) and it will use GPT then automatically. So if this is a problem that should be fixed, then it needs to be fixed now anyway.
@ngompa : This issue focuses on preventing regression, not improving or extending capabilities. Maybe, after resolving the issue a software raid installation is possible on disks > 2 TB, too, as a side effect. But that is not the primary goal.
I just noticed, I mentioned in the test description of the expected results, but not the actual results.
Addendum of the current test results
Test 1 (GPT)
Test 2 (MBR)
Metadata Update from @pboy: - Issue untagged with: meeting
Issue status updated to: Closed (was: Open)
Log in to comment on this ticket.