You are not logged in.
[SYNOPSIS] I somehow had two configs for the same RAID array. Oddly though I don't recall any error messages from mdadm when this happened or while it lasted. I wiped the partition for the second invalid array config and was then able to assign that partition to a new array. The new array was not visible to mdadm --examine --scan until I did this..
I have a few installs booted with syslinux that use the configs below that have been working for some time now but my latest installs with these parameters fail with the subject message.
MODULES=(dm_crypt dm_mod ext2 ext4 nouveau raid0 vfat zram)
HOOKS=(autodetect base block consolefont keymap filesystems keyboard mdadm_udev udev encrypt~cLX hushKernel~cLX LIVE~cLX)
BINARIES=(awk cut date df du grep find hdparm mdmon mkdir pv rsync setfont shuf umount unsquashfs wc zstd)
COMPRESSION=(zstd)Has there been any changes to how the initcpio has to be configured requiring new or different modules, hook, binaries, etc?
Last edited by CaeriTech (2025-09-07 03:15:13)
-=[ LIVE enabled UEFI with redundant syslinux pure systemd detached LUKS header partitionless encrypted GPT SSDx3 RAID0 wayland only because I can. ]=-
Backward compatibility is for the masses. There's no dual-boot here.
[CaeriTech remains only artificially intelligent. Turing would be aghast at just how artificial.]
Offline
What in the world are encrypt~cLX hushKernel~cLX and LIVE~cLX? Even without those, that's a very odd mkinitcpio config, what are you doing?
I can't see how that config could produce that error at all, as the busybox based initramfs should use /new_root, not /real_root, unless it's something strange in those weird hooks.
Last edited by Scimmia (2025-09-05 17:28:35)
Offline
Those are custom hooks that I use to enable detached LUKS header/keyfiles and optionally load the OS to RAM. They have been in use for several years now and I'm trying to use them with new installs but the boot fails now.
-=[ LIVE enabled UEFI with redundant syslinux pure systemd detached LUKS header partitionless encrypted GPT SSDx3 RAID0 wayland only because I can. ]=-
Backward compatibility is for the masses. There's no dual-boot here.
[CaeriTech remains only artificially intelligent. Turing would be aghast at just how artificial.]
Offline
Has there been any changes to how the initcpio has to be configured requiring new or different modules, hook, binaries, etc?
What time frame or version range are you concerned with? Can you not post the custom hooks, full mkinitcpio.conf and the output of mkinitcpio for a build with that config and custom hooks?
Offline
Thanks loqs for suggesting I post the output of mkinitcpio. I decded to run a diff on the output for a working install and it turns out that my custom mdadm.conf wasnt being undated.
The output of this command wasn't being sent to mdadm.conf:
mdadm --examine --scanAs soon as I figure out why I'll update this post to solved status.
Thanks again. You pointed me in the right direction to look to solve this.
[Edit]
It's a bit weird that even when I open the failed install manually from a working install every other array is found by mdadm --examine --scan 4.4-1 but not the array for the failed install. I still have full access to the root of the failed install though.
Last edited by CaeriTech (2025-09-06 15:42:51)
-=[ LIVE enabled UEFI with redundant syslinux pure systemd detached LUKS header partitionless encrypted GPT SSDx3 RAID0 wayland only because I can. ]=-
Backward compatibility is for the masses. There's no dual-boot here.
[CaeriTech remains only artificially intelligent. Turing would be aghast at just how artificial.]
Offline
I ran my install script on a previously booting array WITHOUT re-creating it and it worked because mdadm --examine --scan was able to find it.
So it seems that something is going wrong when I create a new array that makes it invisible to mdadm --examine --scan.
This is the format I use for the command to create the new array with the appropriate device numbers and level of course.
mdadm --create /dev/mdX --level=0 --raid-devices=X --force /dev/sdXAny ideas?
-=[ LIVE enabled UEFI with redundant syslinux pure systemd detached LUKS header partitionless encrypted GPT SSDx3 RAID0 wayland only because I can. ]=-
Backward compatibility is for the masses. There's no dual-boot here.
[CaeriTech remains only artificially intelligent. Turing would be aghast at just how artificial.]
Offline
I think I've found the problem.
The RAID array md/47 has two configurations and one of them has the sdf45 partition assigned to it which I'm attempting to assign to the md/45 ARRAY which does not show under mdadm --examine --scan.
ARRAY /dev/md/47 level=raid0 metadata=1.2 num-devices=4 UUID=0249be04:4008bdb0:2e58af90:c2680d9a
devices=/dev/sdd,/dev/sdc,/dev/sdb,/dev/sda
ARRAY /dev/md/47 level=raid0 metadata=1.2 num-devices=1 UUID=14db9e2f:a2e92191:36050372:84f6cce9
devices=/dev/sdf45How do I remove the md/47 configuration with the sdf45 partition without affecting the md/47 array with the /dev/sdd,/dev/sdc,/dev/sdb,/dev/sda partitions?
Wiping the sdf45 partition using the dd command removed the second invalid config for md/47 and now when I assign sdf45 to md/45 it becomes visible again to mdadm --examine --scan.
Last edited by CaeriTech (2025-09-07 03:05:31)
-=[ LIVE enabled UEFI with redundant syslinux pure systemd detached LUKS header partitionless encrypted GPT SSDx3 RAID0 wayland only because I can. ]=-
Backward compatibility is for the masses. There's no dual-boot here.
[CaeriTech remains only artificially intelligent. Turing would be aghast at just how artificial.]
Offline