In short:

1. modprobe nbd

2. Create the QCOW2 image same (virtual) size as the HD.

3. Connect to the QCOW2 (see link above).

4. Mount the HD's filesystems.

5. On each of these FSen, dd if=/dev/zero of=null.zero bs=1024k; sync; rm null.zero to use up all free blocks, fill them with nulls, then erase the file. Neither perfect nor complete, but it'll zero most free blocks. You will have to <CTRL/C> dd if it doesn't quit when it exhausts free space. (This'll work on all FSen, not just EXT[234].) Also be aware that some FSen may still have limits on the maximum file size; you might have to create several nullX.zero files.

6. Unmount the HD's filesystems.

7. As root, dd --sparse the raw HD device to the QCOW2 NBD device.

8. Disconnect from the QCOW2 (see link above)

9. rmmod nbd

Given the price and size of storage these days (and sufficient funds), it might be better to buy a large SSD or NVME and copy the HD onto a corner of it. If, of course, the above doesn't work well.

Whichever way you use, it will take quite some time. 500GB at 1.5 Gb/s ain't fast. And you're doing it twice (once to zero free blocks, and again to copy the drive).

you can copy the whole disc and then shrink the qcow

I have tried that, but for some reason, errors occurred for me and so I gave it up.

Maybe I am missing something?

dd if=/dev/zero of=/mysparsefile bs=1G count=0 seek=500

Though after that you still have the problem of identifying which blocks on /dev/sda are in use so as to copy those. If you have run zerofree, i.e. made all unused blocks be zeroed, then another dd would do the job, e.g.

dd if=/dev/sda2 of=/mysparsefile conv=sparse,notrunc

Obviously mysparsefile needs to be somewhere with enough space but not on /dev/sda (as it then will fill up the 500G by copying its own blocks)

Yeah, like if you had this situation:

$ df -h
Filesystem      Size  Used Avail Use% Mounted on
/dev/sda1        50G   20G   30G  40% /
/dev/sda2       450G   80G  270G  18% /home

You'd want a single 100GB sparse file, which still retains the 500GB capacity.

Correct, but actually, I may have changed my mind slightly, is there any easy solution in devuan included in doing this? Even that might be helpful. Preferably one that isn't mega slow.

2 hours is fine, anything longer would be painful.

500GB disk btw.

I guess I should try this then?

When creating a partition image file as sparse file, these sectors are omitted anyway.

Then according to its man page,  qemu-img has the -S size argument whereby it translates contiguous runs of zeros as holes.

And dd can do something similar with the conv=sparse argument, though only into a sparse raw image file.

$ df -h
Filesystem      Size  Used Avail Use% Mounted on
/dev/sda1        50G   20G   30G  40% /
/dev/sda2       450G   80G  270G  18% /home

You'd want a single 100GB sparse file, which still retains the 500GB capacity.

zapper wrote:

Anyone know how to copy an entire hard drive onto a qcow2 image but with only the space being used as being counted?

No, but I have spent a significant amount of time trying various solutions that didn't work, before eventually giving up and converting each partition to its own file, with qemu-image convert.

Hmm... that method you mentioned didn't work for me, for some reason.

You know what I mean though, right?

Like say 200GB is used but 450 is available on hdd.

I want only the 200GB to be counted but expansion be possible at the same time.

I am guessing you understand what I am saying, but felt I should clarify anyways. Especially for other people.

I might add to the first post...

Anyone know how to copy an entire hard drive onto a qcow2 image but with only the space being used as being counted?

No, but I have spent a significant amount of time trying various solutions that didn't work, before eventually giving up and converting each partition to its own file, with qemu-image convert.

https://unix.stackexchange.com/question … -browse-re

shows some examples using e2image, also for a qcow2- image. Not sure whether this works for your case too.

Sadly no, but I did find something on that page, that might work.

I will just have to see though...

shows some examples using e2image, also for a qcow2- image. Not sure whether this works for your case too.