1

I know, yet another question about mdadm...

I'll try and be as clear as can be:

System:

Archlinux
Kernel 5.15.81-1-lts
systemd 252.2-3-arch

rescue:

ArchISO (with zfs module)
Kernel 5.10.78-1-lts
systemd 249.6-3-arch
Zfs 2.1.1-1

My system consists of two drives:

nvme{0,1}n1:
- p1 - F32 - md1 (Metadata=1.0, homehost=any)
- p2 - zfs - root fs
- p3 - luks - md3 (Metadata 1.0, homehost=any)

The Idea is as follows:

  1. EFI detect either of the two first partition (Metadata 1.0 implies that raid data are at the end of the partition)
  2. EFI start, calling initramfs
  3. Udev rules should (did) assemble the different raid
  4. Luks decipher the md3 and mount it.
  5. Zfs-mount.service loads keys and mount fs
  6. System is happily up and running

In rescue, I can assemble the raid, decipher and access everything.

I went and extracted initramfs, to check all udev rules.

I could assemble the raid using udev's rules's command line (mdadm --incremental --export $devnode)

I am not sur why he would ignore the devices (or even if it actually ignored them).

Though, all of it was working a day ago, at most...

FYI: the raids where both created like this:

$name and $part are respectively boot/1 and secure/3:

mdadm --create --metadata=1.0 --name=$name --raid-device=2 --homehost=any /dev/md$part /dev/nvme{0,1}n1p$part

I also tried creating a manual hook that simply assembled both raids it was checking for mdadm.conf and didn't seem to work (written from memory, I'll edit it asap, edited):

run_hook(){
    if [ -r /etc/mdadm.conf ] && [ -x /usr/bin/mdadm ]; then
        /usr/bin/mdadm -A /dev/md1 -c /etc/mdadm.conf
        /usr/bin/mdadm -A /dev/md3 -c /etc/mdadm.conf
    fi
}

I'll work on it later on (3:00 am...) and correct / complete as much info.

If you have any idea of where should i look for anything wrong ?

I'll provide as much info i can , but i only have my phone...

Thanks in advance

0

You must log in to answer this question.

Browse other questions tagged .