Using Ubuntu Cloud Image 20.04 on CentOS 8 Cockpit KVM

To use ubuntu cloud image on CentOS 8 Cockpit KVM, we would need to download the daily image

There are many image format, but I only tested 2 of them, 

• https://cloud-images.ubuntu.com/focal/current/focal-server-cloudimg-amd64-disk-kvm.img, this is not booting
• https://cloud-images.ubuntu.com/focal/current/focal-server-cloudimg-amd64.img, this boots fine

$ wget https://cloud-images.ubuntu.com/focal/current/focal-server-cloudimg-amd64.img

After downloading we can verify that this is in qcow format by using

$ file focal-server-cloudimg-amd64.img 

focal-server-cloudimg-amd64.img: QEMU QCOW Image (v2), 2361393152 bytes

or

$ qemu-img info focal-server-cloudimg-amd64.img

image: focal-server-cloudimg-amd64.img

file format: qcow2

virtual size: 2.2G (2361393152 bytes)

disk size: 522M

cluster_size: 65536

Format specific information:

    compat: 0.10

    refcount bits: 16

Let's resize the image, since 2.2 GB looks a bit too small for me

$ qemu-img resize focal-server-cloudimg-amd64.img 15G

You can verify the result of resize by running qemu-img info again.

By default, this image expect cloud-init operation. We will try to bypass that.

First we will resize the partition table to use the extra space allocated from resize operation before, then we mount the image for further adjustment.

$ sudo modprobe nbd

$ sudo qemu-nbd -c /dev/nbd0 focal-server-cloudimg-amd64.img

$ sudo blkid /dev/nbd0p1

/dev/nbd0p1: LABEL="cloudimg-rootfs" UUID="9be2d018-238c-4f91-a3a2-9041df4c3f47" TYPE="ext4" PARTUUID="old-uuid"

we will delete partition 1 and recreate it

$ sudo fdisk /dev/nbd0

(fdisk) p

note the partition start and end

(fdisk) d

delete partition 1

(fdisk) n

create partition 1 again with the same start sector and updated end sector

(fdisk) w

there will be warning about disk signature, leave them as is, and exit fdisk

since we recreate partition 1, the PARTUUID will change and we need to adjust grub.cfg

$ sudo blkid /dev/nbd0p1

/dev/nbd0p1: LABEL="cloudimg-rootfs" UUID="9be2d018-238c-4f91-a3a2-9041df4c3f47" TYPE="ext4" PARTUUID="new-uuid"

note down this new uuid, we will need it later

$ ls /image || mkdir /image

$ sudo mount /dev/nbd0p1 /image/

Then we chroot to the new environment and change root password, so we can login to the VM later

$ sudo chroot /image/

# passwd

set your new password

The default grub.cfg also require serial port to be available, otherwise it will freeze while booting. See here for the discussion https://bugs.launchpad.net/cloud-images/+bug/1573095

We can disable serial port requirement by editing grub.cfg and removing all reference to console=ttyS0. Also edit file /etc/default/grub.d/50-cloudimg-settings.cfg

Do not forget to change all reference to PARTUUID from the old uuid to the new uuid. The following files need to be changed

/boot/grub/grub.cfg

/etc/default/grub.d/40-force-partuuid.cfg

Exit the chroot environment and umount everything

# exit

$ sudo umount /image/

$ sudo qemu-nbd -d /dev/nbd0

Import VM from Cockpit interface, we can use virtio as the disk interface without issue. If cockpit do not list Ubuntu 20.04 as OS choice, we can use the nearest Ubuntu. In my case I am using Ubuntu 19.10.

Test your VM by login as root in the console interface. 

To enable ssh, you have to generate hostkey and enable password authentication in sshd_config

# ssh-keygen -A

in sshd_config

PasswordAuthentication yes

# if needed

PermitRootLogin yes

Enjoy !!