Am 09.06.2023 um 18:19 schrieb François Patte
<francois.patte(a)mi.parisdescartes.fr>:
Sorry for the delay.
Le 2023-06-09 14:01, Peter Boy a écrit :
>> Am 09.06.2023 um 09:48 schrieb François Patte
<francois.patte(a)mi.parisdescartes.fr>:
>>> ...
>>> Is it possible to unplug one of the disks (say sdf) and make an
>>> install with a raid1 in degraded mode (+lvm) then, when the install is
>>> finished, to plug again the disk sdf and synchronise the partitions
>>> with mdadm?
>>> Moreover, I don't understand why the problem occurs only with sde1 and
>>> sdf1 and not with the other disks (sde2, sdf2 as already pointed) : I
>>> have 6 disks, and more partitions, paired in pairs as raid members.
>>> Thank you for lights.
>> No answer... Is this problem so uncommon? Or so obvious?
> I suppose, the issue is a bit complex and a bit confusing.
> The first question is which Fedora edition you want to install. The
> installation routines differ a lot and so do the possible solutions.
I tried f38 netinstall.
Fedora Server Edition is the only Edition with a net install media, as far as I know.
Nevertheless, do you want to install (and operate) a Fedora Server or just use the medium
to install another variant (by selecting a different target in the installation source
menu). These alternatives require different followup steps.
> Additionally, in the previous thread, I recall that the affected
disks
> are to be reinstalled any way and can be reformatted beforehand. Is
> that not a solution?
Reformat the disks where the system is installed (2 on 6) OK, but why should I reformat
the other disks?
No, just reformat the disk(s) which cause issues. And probably only reformat the
partition(s) that are affected.
And I don't understand why there are duplicates UUID and why the
f38 installer complains about these duplicates UUID on disk /dev/ssde1 and /dev/sdf1 and
not for the duplicates UUID on /dev/sde2 and /dev/sdf2?
It’s hard to identify the cause of the problem. But the problem is not necessarily the
identical UUIDs in that specific field, but probably a value anywhere else that determine
the purpose of the disk resp. the partition, in your case usage of the partition as a raid
member and not as an independent partition (which would require unique UUIDs).
You can try to disconnect one of the drives of the failing raid and then reboot. If it
works, fine. Otherwise try the other disks. If the system boots and works without issues
(besides a degraded raid) with a one-disk-raid, then you may delete all partitions of the
disconnected disk, probably even delete the partition table (e.g. by using dd if:/dev/zero
of=/dev/<disconneted disk>). Then integrate the empty disk into the raid again and
initiate a resynchronization. It may be a bit tricky.
If you can’t manage to resync the raid, you have at least still a bootable system and a
working base for next steps.
The alternative is to reformat either the disk or just one or two partitions (that depends
on the details of your disk layout) and then start a complete new installation of the
system (and preserving user data if your existing partitioning scheme allows for that).
The details here depend on what Edition you really want to install.
--
Peter Boy
https://fedoraproject.org/wiki/User:Pboy
PBoy(a)fedoraproject.org
Timezone: CET (UTC+1) / CEST /UTC+2)
Fedora Server Edition Working Group member
Fedora Docs team contributor and board member
Java developer and enthusiast