Re: No fool like an old fool (debian installation probs)
On Mon 01 May 2023 at 03:11:56 (+0200), DdB wrote:
> For example, when i allow the use of the swap partition, it gets
> reformatted and another PARTUUID is assigned to it, leading to failures
> booting other systems from that disk.
The easy way to avoid that problem is to use LABEL rather than
(PART)UUID, because any time after the partitioner has run, while
it's installing more of the system, you can reset the LABEL back
to how you had it.
# /target/sbin/swaplabel -L whatever /dev/sdXn
You can set the LABEL for the new system at leisure, because the
UUID that the installer used will still be working.
(I parenthesised (PART)UUID because I thought the installer set
and used UUIDs, not PARTUUIDs; but honestly, I might not notice if
any PARTUUIDs got reset: I only use LABELs and PARTLABELs myself.)
> OTOH Disallowing its use for the
> time being leads to install failures due to the 4GB of RAM apparently
> not being enough for that VM.
I can install bullseye and bookworm-RC1 standard software (no DE) on
an i386 laptop with ½GB memory using no swap, so that surprises me.
> Another problem, similar but distinctly different happens around the OS
> partition. Whenever i gave it to the installer, it reformatted it while
> changing the PARTUUID, again leading to failures in the boot process of
> the other partitions... Until i discovered this workaround: Assign the
> partition to ext4 and the mountpoint / but not allowing to delete its
> content (as that would induce the reformatting). Instead, after
> finishing partitioning, i asked for a shell and from there manually
> deleted ONLY THE DATA from the partition(s) used. Then installation
> progressed further.
I presume your setupp (with VMs etc) is much more complex that my own,
so I can't understand why the booting process of one partition would
be affected by not identifying a different system on the same machine…
> BTW: Apparently, this was the only way to keep the PARTUUID, as deleting
> the data by overwriting with zeroes lead to the installer overwriting
> the fs structures making another mke2fs necessary.
… and I don't know which data you would be zeroing.
> But the worst of all, was the software selection: Even though i asked
> for debconf priority low, i got only a few checkboxes to pick from. And
> choosing GNOME did automatically install software i neither need nor want.
>
> But omitting GNOME from the list lead to a system failing to boot with
> tons of messages stating the absense of all kind of gnome parts.
>
> So, just in order to make some progress, i had to willingly and
> temporarily kill the bootability of other partitions in order to have
> the install terminate without failures.
>
> At that point, i decided to try the netinstall, as it is now available
> with release 11.7 and i did expect there to be more choices. And when
> even that one failed to boot without GNOME, i even did check the
> sha512sum to make sure, i was using the proper one. - I did!
Yes, sorry, I don't use DEs. You might be able to find discussions
in the past list archives about what's installed when you select
Debian desktop environment without any DE from the list below.
> Now, i am confused. Many years ago, the expert installer was very
> difficult to use and did provide very many choices (and liberties to
> spoil the fun). Now, i dont seem to be able at coming to the stage,
> where i can seriously test more complicated things (like multi-boot,
> zfs, and other things) that make up my current host in its buster version.
That sounds as if you're harking back to the days when part two of
installation was running dselect after a reboot. That was tedious.
Since etch, it's been very much more straightforward.
> It is quite obvious, that i am lacking SOME kind of understanding,
> because what i am trying to accomplish should really be quite easy, but
> i am only experiencing troubles.
I think you might have to explain more about the overall configuration
of your systems: I find it difficult to see how these booting
processes interact with one another. I have PCs with up to three
systems on them (buster/bullseye/bookworm), sharing /home and swap
(and obviously ESP), but they depend on each other beyond the fact
that each will mount the others' root filesystems (readonly) as a
convenience.
> Oh, as a side note: As a nicety of my setup, i can use bash scripts,
> that work well from my main OS, or from the simulation VM which can be
> used to test functionality before it is rolled out to the host. That is
> why, i am relying/insisting on absolutely identical PARTUUIDs inside
> that VM.
(PART)LABELs instead?
> Now, i am really curious as to how to get to a proper system without
> having to go through the baby-steps beginning with debootstrap, and all
> the millions of configuration steps, that i did not even bother to list.
>
> I am really interested to understand, in which way i am creating all
> those problems for myself. Could it be related to the fact, that i am
> accessing the net fom a VPN? - I don't think so.
I wouldn't think so either. When I install, I invariably use a proxy
except on the machine the does the proxying.
> Any suggestions/questions/hints from the power-users in here?
> ... would be wildly appreciated ...
Only (PART)LABELs instead, I'm afraid. Not a power-user.
Cheers,
David.
Reply to: