Error partitioning with RAID

[Diego Zuccato]

Il 05/03/2025 11:59, Thomas Lange ha scritto:

> I never tried such a configuration and I don't know if this can work.
Didn't think that having to keep some RAID volumes across reinstalls 
would be so problematic :) Should be quite a common scenario :)

> Can you please check if the RAID is already assembled before
> setup-storage starts? I'm not sure if setup-storage can handle an
> already running raid, or if it wan't to detect the partitions and the
> reassembles the RAID itself.
The RAID is not assembled, or at least when setup-storage aborts install 
there is no /proc/mdstat .

> There's a kernel cmdline option for dracut, which boots the FAI
> environment (option callled rd.auto=1) which can enable building the
> RAIDs.
Currently not using it.

We're still doing some tests, but this config seems to work, at least if 
disks are wiped when booting with 'initial' flag:

-8<--
disk_config disk1 bootable:1 align-at:1M preserve_lazy:3,6,7
primary -       200M         -       -
primary swap    2G           swap    -
primary -       120M-128M    -       -
logical -       15G          -       -
logical -       500G         -       -
logical -       1G-          -       -

disk_config disk2 sameas:disk1
disk_config disk3 sameas:disk1

# preserve_lazy references md devices only, starting from md0 !
disk_config raid fstabkey:uuid preserve_lazy:1,3,4
raid1   /boot           disk1.1,disk2.1,disk3.1   ext4 
rw,noatime,errors=remount-ro mdcreateopts="--metadata=0.90"
raid1   -               disk1.3,disk2.3,disk3.3   ext4  noauto,noatime 
createopts="-L PERMDATA"
raid5   /               disk1.5,disk2.5,disk3.5   ext4 
rw,noatime,errors=remount-ro
raid1   /var/lib/mysql  disk1.6,disk2.6,disk3.6   ext4 
rw,noatime,nosuid,nodev
raid5   /home           disk1.7,disk2.7,disk3.7   ext4 
rw,noatime,nosuid,nodev
-8<--

As noted in the comment, preserve_lazy references different objects 
(partitions in the first disk_config, md devices in the second). The 
worst thing is that it seems it's not allowed to specify preserve_lazy 
after the names of the md device components (diskX.Y...:preserve_lazy).

Some tests on a VM were OK, and preserved partitions did not get erased. 
I'm still not sure why I have to wipe the disks even when using 
flag_initial and why preserve_reinstall errors out when disks are wiped 
(IIUC if flag_initial=1 it shouldn't even try to read partitions info....).

-- 
Diego Zuccato
DIFA - Dip. di Fisica e Astronomia
Servizi Informatici
Alma Mater Studiorum - Università di Bologna
V.le Berti-Pichat 6/2 - 40127 Bologna - Italy
tel.: +39 051 20 95786