Short guide to preparing VMs for local development

Trivial guide for setting up cloud images (incl. Ubuntu Core) for local development in VM.

It's assumed that the images are configured to run cloud-init.

Local cloud-init config

Example cloud configuration file:

#cloud-config

datasource_list: [ NoCloud, None ]
# default user password, does not work?
password: guest
chpasswd:
  expire: False

users:
  - name: guest
    # 'guest', does not work?
    passwd: $1$xyz$NupBwZXNoMXD8NQwzjRW/0
    # for sudo
    groups: wheel
    # sudo without password
    sudo: ALL=(ALL) NOPASSWD:ALL
    # eplicitly set the shell
    shell: /bin/bash
    ssh-authorized-keys:
      - ssh-rsa AAAAB3NzaC1yc2EAAAABIwAAAQEA1LCRbWnZ/GmM+LZex06HjNyw0aixYbD3P8mfsNHcBV6/LNk7vxw7+5nhooaBlkv1X/hpc/q3BnGy2W4goJ8aEL3JliJl/+4ijdSEwYXKZj0PpFPY8ir4VOFzVlPIX4SoHgheSo5it7zFRBpHMtDqSmgoWmzFKLX3qg144Cv9Lxkqpkx0ndpflYsLz8hH9WT85OjNvVI51lxoTq86XmU0rSzQkT0vpNwGRHSs0HS197d4ym9f1dGFTflWaKmhrquvBWTztHvfhxgZz6OdEeyywAdQBBi3sbWkRuwjZ/aX5K3obwIJL0iJN8hwf64Wt5plYQVhIrNlqVCU8ZRLuzLAGw== maciek@corsair

# for Ubuntu Core
snap:
  commands:
    # force add the user
    00: snap create-user --sudoer --force-managed [email protected]

# get a debug shell
bootcmd:
    - systemctl enable --now debug-shell.service

Build the ISO image:

$ cloud-localds uci-data-guest.img uci-data-guest

Preparing Ubuntu Core

UC20 model

Grab the model definition (UC20 model requires snap 2.43+)

$ snap known --remote model model=ubuntu-core-20-amd64 series=16 brand-id=canonical > pc-20.model

Build image

$ ubuntu-image snap pc-20.model -w $PWD/image-home [--extra-snaps <mysnap>.snap] [--image-size=10G]

⚠️ For Ubuntu Core 20, the image must be at least 10GB (see --image-size=10G).

The output image and some intermediate data ends up in $PWD/image-home

Run VM

Example UC20 VM:

$ qemu-system-x86_64 -enable-kvm \
    -snapshot \
    -m 2048 \
    -smp 4 \
    -device virtio-net-pci,netdev=mynet0 -netdev user,id=mynet0,hostfwd=tcp:127.0.0.1:59444-:22 \
    -serial telnet:127.0.0.1:59483,server,nowait \
    -monitor telnet:127.0.0.1:59426,server,nowait \
    -object rng-random,filename=/dev/urandom,id=rng0 -device virtio-rng-pci,rng=rng0 \
    -drive file=./image-home/pc.img,if=virtio,index=0 \
    -drive file=uci-data-guest.img,if=virtio,index=1,snapshot=on \
    -bios /usr/share/ovmf/x64/OVMF_CODE.fd

Command details:

4 CPUs, 2GB RAM
snapshot mode (no permanent changes to disk images)
SSH port forwarded to local 59444
serial on port 59483
qemu monitor on port 59426
virtio RNG
virtio network interfaces
virtio disk access
optional UEFI support

In snapshot mode, qemu uses $TMPDIR (usually /tmp) to create a temporary file for a new disk image, and then unlinks it. If your $TMPDIR is on tmpfs you may run out of RAM.

Building custom CentOS cloud image

The instructions should work for RHEL and Fedora too.

CentOS, RHEL and Fedora use Anaconda for installation, you can feed a custom kickstart file named ks.cfg on a volume named OEMDRV.

Grab the CentOS cloud SIG image build tools:

$ git clone http://github.com/CentOS/sig-cloud-instance-build
$ cd sig-cloud-instance-build/cloudimg

Use one of the existing kickstart files and edit it to your liking:

$ mkdir floppy
$ cp CentOS-8-x86_64-hvm.ks floppy/ks.cfg
$ vim floppy/ks.cfg

Build the disk image:

$ mkfs.ext2 -d floppy -L OEMDRV floppy.img

Prepare the image file:

$ qemu-img create -f qcow2 centos-cloud.img 20G

Grab the network install CD of CentOS. Run the VM and wait for it to power down:

$ qemu-system-x86_64 -enable-kvm \
    -m 2048 \
    -smp 4 \
    -device virtio-net-pci,netdev=mynet0 -netdev user,id=mynet0,hostfwd=tcp:127.0.0.1:59444-:22 \
    -serial telnet:127.0.0.1:59483,server,nowait \
    -monitor telnet:127.0.0.1:59426,server,nowait \
    -object rng-random,filename=/dev/urandom,id=rng0 -device virtio-rng-pci,rng=rng0 \
    -drive file=centos-cloud.img,if=virtio,index=0 \
    -drive file=floppy.img,if=virtio,index=1,snapshot=on \
    -cdrom CentOS-8-x86_64-1905-boot.iso