** DONE Automating VM Setup with cloud-init on TrueNAS Scale :blog:admin:tech:cloud:
CLOSED: [2023-12-19 Tue 11:58]
:PROPERTIES:
:ID:       truenas-automatic-vm-cloud-init
:CREATED:  [2023-12-19 Tue]
:END:
:LOGBOOK:
- State "DONE"       from "TODO"       [2023-12-19 Tue 11:58]
:END:
I hate repeated manual labor.
So I set out to use my experience gained with cloud-init in experimenting with [[https://harvesterhci.io/][Harvester]]
to automate the setup of Virtual Machines with the help of
https://blog.robertorosario.com/setting-up-a-vm-on-truenas-scale-using-cloud-init/.

A few notable things to mention:
- I created a child dataset in each of our datasets
  ([f]ast=SDD for root filesystem, [b]igdata=HDD for storage)
  specifically for VMs, this makes administration much easier
- I went with debian,
  which has a slim genericcloud image,
  which worked on Harvester
  but it does not work here
  because the cloud-init config is mounted as CD-ROM -
  so after some tinkering I found the generic image to work well
- I added the password, which is now unfortunately mandatory

The section "BOTH" contains a few variable definitions
that should be provided before every of the other commands.
The "LOCAL" section exists because the ~cloud-image-utils~ package is needed
and TrueNAS Scale does not allow you to easily install packages.
It will presume you provided a suitable cloud-init config in =<VM-Name>-seed.qcow.yaml=
and then convert and copy it over to the TrueNAS Scale server
(which I shorthanded as =nas= in my ssh config).
All other variables should be self-explanatory.
I did not extract the size of the disks into variables
because we usually just go with lavish defaults
due to our setup having plenty of space (>100TB).

Here are the commands in a compacted fashion,
for easy copying out:

#+begin_src sh
# LOCAL

${EDITOR:-nano} $SEEDFILE.yaml
cloud-localds --verbose $SEEDFILE $SEEDFILE.yaml
scp $SEEDFILE nas:${IMAGE_PATH}

# BOTH

VM_NAME=nostr
IMAGE_PATH=/mnt/b/media/iso/servers/
SEEDFILE=${VM_NAME}-seed.qcow2

# REMOTE

VM_PATH=eagle/vm/${VM_NAME}
 # do not use genericcloud here as it is missing CDROM drivers
VM_IMAGE=http://cloud.debian.org/images/cloud/bookworm/latest/debian-12-generic-amd64.qcow2
 # extra data volume, comment out to omit
VM_DATA=b/vm/${VM_NAME}-data
VM_MEMORY=$(expr 8 \* 1024)
VM_PASSWORD=password

sudo zfs create -V 40G "${VM_PATH}"
test -n "${VM_DATA}" && sudo zfs create -V 150G "${VM_DATA}"
cd "${IMAGE_PATH}"
test -e "$(basename ${VM_IMAGE})" || wget "${VM_IMAGE}"
case "${VM_IMAGE}" in
(*.raw) sudo dd if=$(basename ${VM_IMAGE}) of=/dev/zvol/${VM_PATH} status=progress bs=1M;;
(*) sudo qemu-img convert -O raw $(basename ${VM_IMAGE}) /dev/zvol/${VM_PATH};;
esac

# Create the VM
RESULT=`midclt call vm.create '{"name": "'${VM_NAME}'", "cpu_mode": "HOST-MODEL", "bootloader": "UEFI_CSM", "cores": 2, "threads": 2, "memory": '${VM_MEMORY}'}'`
VM_ID=`echo ${RESULT} | jq '.id'`
# Add the display
midclt call vm.device.create '{"vm": "'${VM_ID}'", "dtype": "DISPLAY", "order": 1004, "attributes": {"web": true, "type": "VNC", "bind": "0.0.0.0", "password": "'${VM_PASSWORD}'", "wait": false}}'
# Obtain a random MAC address
MAC_ADDRESS=`midclt call vm.random_mac`
# Add the NIC
midclt call vm.device.create '{"vm": "'${VM_ID}'", "dtype": "NIC", "order": 1010, "attributes": {"type": "VIRTIO", "nic_attach": "br0", "mac": "'${MAC_ADDRESS}'"}}'
# Add the root disk
midclt call vm.device.create '{"vm": "'${VM_ID}'", "dtype": "DISK", "order": 1001, "attributes": {"path": "/dev/zvol/'${VM_PATH}'","type": "VIRTIO"}}'
# Add a data disk
midclt call vm.device.create '{"vm": "'${VM_ID}'", "dtype": "DISK", "order": 1002, "attributes": {"path": "/dev/zvol/'${VM_DATA}'","type": "VIRTIO"}}'
# Add the CDROM
midclt call vm.device.create '{"vm": "'${VM_ID}'", "dtype": "CDROM", "order": 1005, "attributes": {"path":"'${IMAGE_PATH}${SEEDFILE}'"}}'
#+end_src