TRIM with linux guest for minimal img file

[motophil]

Hi all,

I've found this and subsequently this thread, but nothing about linux guests.

I would like to have my img files as small as possible (of course), so I tried to set it up as an SSD similar to Windows.
This seems to work now, but I'm not very experienced with this stuff so this is a request for improvements as well as a how-to for others with similar little experience.

When creating the VM, I chose:
Primary vDisk bus: SCSI
Primary vDisk type: qcow2 (if I choose raw here the img file will be full size even before first boot)
Manually edited the VM XML file and added discard='unmap'
 

    <disk type='file' device='disk'>
      <driver name='qemu' type='qcow2' cache='writeback' discard='unmap'/>
      <source file='/mnt/user/domains/Geraldine/vdisk1.img' index='2'/>
      <backingStore/>
      <target dev='hdc' bus='scsi'/>
      <serial>vdisk1</serial>
      <boot order='1'/>
      <alias name='scsi0-0-0-2'/>
      <address type='drive' controller='0' bus='0' target='0' unit='2'/>
    </disk>

Now I installed ubuntu server minimal. This tells me to check lsblk --discard to see if my (virtual) drive supports trim, so I did:
 

phil@geraldine:~$ lsblk --discard
NAME                      DISC-ALN DISC-GRAN DISC-MAX DISC-ZERO
sda                              0        4K       1G         0
├─sda1                           0        4K       1G         0
├─sda2                           0        4K       1G         0
└─sda3                           0        4K       1G         0
  └─ubuntu--vg-ubuntu--lv        0        4K       1G         0
sr0                              0        0B       0B         0

So the DISC-GRAN and DISC-MAX beeing non-zero tells me it does.

Also, lvm.conf has issue_discards = 1
 

phil@geraldine:~$ fgrep issue_discards /etc/lvm/lvm.conf
        # Configuration option devices/issue_discards.
        issue_discards = 1

But my /etc/fstab knows nothing about discard:
 

phil@geraldine:~$ fgrep discard /etc/fstab
phil@geraldine:~$

 

so i added it manually
 

phil@geraldine:~$ fgrep discard /etc/fstab
/dev/disk/by-id/dm-uuid-LVM-kcbZKFWiEPoidmRd4HaCFzVI2y7GhD9QHQdKDxrBf5cDOvyjd4PTF0iX0wfApBXL / ext4 defaults,discard 0 1
/dev/disk/by-uuid/84b65076-9b7c-4efe-8ce8-06d5688c737f /boot ext4 defaults,discard 0 1
/dev/disk/by-uuid/0E6B-E392 /boot/efi vfat defaults,discard 0 1
phil@geraldine:~$

 

also did update-initramfs -c -k $(uname -r) because I wasn't sure if I had to

and rebooted. Now the discs are mounted with the discard option.
 

phil@geraldine:~$ mount | grep discard
/dev/mapper/ubuntu--vg-ubuntu--lv on / type ext4 (rw,relatime,discard)
/dev/sda2 on /boot type ext4 (rw,relatime,discard)
/dev/sda1 on /boot/efi type vfat (rw,relatime,fmask=0022,dmask=0022,codepage=437,iocharset=iso8859-1,shortname=mixed,errors=remount-ro,discard)

Now hdparm still does not mention TRIM:
 

phil@geraldine:~$ sudo hdparm -I /dev/sda

/dev/sda:

ATA device, with non-removable media
Standards:
        Likely used: 1
Configuration:
        Logical         max     current
        cylinders       0       0
        heads           0       0
        sectors/track   0       0
        --
        Logical/Physical Sector size:           512 bytes
        device size with M = 1024*1024:           0 MBytes
        device size with M = 1000*1000:           0 MBytes
        cache/buffer size  = unknown
Capabilities:
        IORDY not likely
        Cannot perform double-word IO
        R/W multiple sector transfer: not supported
        DMA: not supported
        PIO: pio0

But it seems to be working:
 

phil@geraldine:~$ sudo fstrim -A -v
/boot/efi: 1 GiB (1118625792 bytes) trimmed on /dev/disk/by-uuid/0E6B-E392
/boot: 1.8 GiB (1898135552 bytes) trimmed on /dev/disk/by-uuid/84b65076-9b7c-4efe-8ce8-06d5688c737f
/: 54.7 GiB (58719956992 bytes) trimmed on /dev/disk/by-id/dm-uuid-LVM-kcbZKFWiEPoidmRd4HaCFzVI2y7GhD9QHQdKDxrBf5cDOvyjd4PTF0iX0wfApBXL

So, please feel free to comment. Thank you.

 

[motophil]

I learned that kvm does not remove the space freed by trim from the img file automatically. This can be done with the qemu-img convert command.

 

qemu-img convert -c -O qcow2 input.img output.img

 

This converts a disk image into another format - in this case, qcow2 format. Since input is qcow2 format already it is not converted, just copied - but empty space is omitted from the copy. The -c option also enables compression, which slows down access to the image but saves a lot of space.
 

So to clean up and compress your image, power down your VM, open a root console, cd into the matching directory (usually /mnt/user/<VMName>) and do the following:
 

mv vdisk1.img vdisk1.img.bak
qemu-img convert -c -O qcow2 vdisk1.img.bak vdisk1.img
chown nobody vdisk1.img
chgrp users vdisk1.img
chmod 777 vdisk1.img

 

If your VM starts up fine, you can delete the vdisk.img.bak file.

 

 

[JonathanM]

On 1/8/2024 at 4:24 AM, motophil said:

(if I choose raw here the img file will be full size even before first boot)

It only appears that way in one view. It's a sparse file, meaning the full capacity is shown, but only the in use sections are allocated and take up space. You can easily have multiple 250GB sparse files on a single 250GB disk, but you will run into issues when more than 250GB worth of data is actually used in any combination of the files.

[aim60]

Try these switches to the ls command

 

 ls -lsh
total 151M
65M -rwxrwxrwx 1 nobody users 20G Dec 30 15:02 vdisk1.img*
86M -rwxrwxrwx 1 nobody users 10G Dec 30 15:02 vdisk2.img*

 

The size on the left is the allocated space.