IBM Cloud cluster-api: building a CAPI image

Per the IBM Cloud Kubernetes cluster-api provider, I followed the raw instructions with some amendments.

Steps

1. Provision an Ubuntu 20.04 image.

2. Update the apt repository

$ apt update

3. Install the dependencies (more than what’s in the instructions)

$ apt install qemu-kvm libvirt-daemon-system libvirt-clients virtinst cpu-checker libguestfs-tools libosinfo-bin make git unzip ansible python3-pip

4. Clone the image-builder repo

$ git clone https://github.com/kubernetes-sigs/image-builder.git

5. Change to the capi image

$ cd image-builder/images/capi

6. Make the deps-raw to confirm everything is working.

$ make deps-raw

7. Create the ubuntu-2004 image.

$ make build-qemu-ubuntu-2004

Once complete you’ll see:

==> qemu: Running post-processor: custom-post-processor (type shell-local)
==> qemu (shell-local): Running local shell script: /tmp/packer-shell078717884
Build 'qemu' finished after 12 minutes 8 seconds.

==> Wait completed after 12 minutes 8 seconds

==> Builds finished. The artifacts of successful builds are:
--> qemu: VM files in directory: ./output/ubuntu-2004-kube-v1.22.9
--> qemu: VM files in directory: ./output/ubuntu-2004-kube-v1.22.9

8. Append the .qcow2 extension

$ mv ./output/ubuntu-2004-kube-v1.22.9/ubuntu-2004-kube-v1.22.9 ./output/ubuntu-2004-kube-v1.22.9/ubuntu-2004-kube-v1.22.9.qcow2

You can now upload the output to IBM Cloud Object Storage.

A couple quick tips:

• If you see any warnings, you can get advanced details using export PACKER_LOG=1 which puts out the full packer logging. see Packer
• KVM module not found indicates you are running in a nested KVM, you’ll have to swap out of the VM and enable nested KVM. Fedora: Docs
• Adding a VM to VPC is documented here Console: customImage