On 14.01.2024 13:15, Andy Smith wrote:
On Sun, Jan 14, 2024 at 08:01:52AM +0000, Andy Smith wrote:
If necessary and if there is a way, I *can* nuke off the target
machine's "foo" volume group and recreate the RAID array if I have
to make it 512e format. But obviously I'd like some way to move this
disk image and have it still work without having to meddle inside it
much — it is a VM disk.
I think I may be able to use hdparm to reformat these Ultrastar DCs
from 4kn to 512e…
hdparm --set-sector-size 512 --please-destroy-my-drive /dev/sdX
https://wiki.archlinux.org/title/Advanced_Format#Advanced_Format_hard_disk_drives
It is okay to destroy the drive contents as there's nothing working
on there yet given this initial failure.
I think that is the only way this is going to work without mounting
filesystems both sides and using an fs-aware tool like tar|ssh or
rsync… which I don't want to do.
Thoughts?
Weird. I can imagine some wild theories. :)
I think it has something to do with a fact that "disk_image" is a LVM
volume and you dd copy "/dev/mapper/VG--name-LV--name" not "/dev/sdc2" .
Could it be that "dd" sees source and destination as block devices and
copies data by sector? I.e. takes 512b sector and place it to the
beginning of 4096b sector and padding the rest 3584b of the sector with
zeroes.
This calls for some tests. I'd create 1G image test file with dos
partition table and a few primary partitions:
$ dd if=/dev/zero of=./1G-disk-image.bin bs=1M count=1000
1000+0 records in
1000+0 records out
1048576000 bytes (1,0 GB, 1000 MiB) copied, 8,07768 s, 130 MB/s
...
$ /sbin/fdisk -l ./1G-disk-image.bin
Disk ./1G-disk-image.bin: 1000 MiB, 1048576000 bytes, 2048000 sectors
Units: sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disklabel type: dos
Disk identifier: 0x143acb59
Device Boot Start End Sectors Size Id Type
./1G-disk-image.bin1 2048 411647 409600 200M 83 Linux
./1G-disk-image.bin2 411648 1435647 1024000 500M 83 Linux
./1G-disk-image.bin3 1435648 2047999 612352 299M 83 Linux
And dd copied it the same way to disk with 4k sectors.
Just to see if it will be inconsistent too, without LVM mapper involved.
I'd also try to clone disk to image file first, with say "partclone"
utility, "scp" the image file and restore it at remote destination with
the same utility after recreating dumped partition table layout using
"sfdisk".
I did this procedure multiple times successfully, but without 4k sector
disks and LVM involvement.
--
With kindest regards, Alexander.
⢀⣴⠾⠻⢶⣦⠀
⣾⠁⢠⠒⠀⣿⡁ Debian - The universal operating system
⢿⡄⠘⠷⠚⠋⠀ https://www.debian.org
⠈⠳⣄⠀⠀⠀⠀
