At Tue, 7 Oct 2025 14:48:48 -0500 Ram Ramesh <rramesh2400@gmail.com> wrote:
Yes what you've described will work except for having to shrink the
second partition a little for the backup GPT header. mdadm guides have
got you there; should be pretty simple for a RAID-1 that's not totally
full.
As I have indicated, I am deleting everything in the disk as I convert
to GPT with same size MD partition. I have about 32G of free space to
work with as the current MBR partition 1 is empty/unused. So, there
should be no issues with secondary table at the end. The new partitions
will automatically move up a bit and RAID rebuild will take care of
initializing the new partition after I clobber the disk and recreate a
new GPT layout.
Most people seem to opt for having two ESPs that are kept in sync and
thus both/either are selectable from the firmware boot process.
I am not too worried about any redundancy on EFI/boot partitions as a
broken EFI is easy to fix with booting install disk in rescue mode and
reinstalling grup again.
What you should do is make EFI/boot partitions on both disks. And then
manually sync them -- this is not a "continious" process, just when something
changes (eg new kernel, new EFI settings, whatever). If the "boot" disk fails,
the other has a copy of the EFI stuff, so you don't need to mess with a rescue
disk, just swap the second disk to the boot disk position, and the system will
boot (with the RAID in degraded mode). When a replacement disk arrives,
install it as the second disk, partition it, sync over the EFI/boot
partituons, and do the usual RAID rebuild for the main data partiition.
Regards
Ramesh