This repo contains files and instructions for building customized RHCOS (Red Hat CoreOS) disk images that are used for installation and bootstrapping of OpenShift Clusters.
Some background context and some examples for creating layered RHCOS container imags can be found in the OpenShift Documentation. Some of that is reproduced here to provide a full example.
For this to work you will need a registry pull secret. If you have a cluster up and running already then you most likely have that set up. If not, then you should be log in and grab your pull secret from console.redhat.com.
In order to figure out what container image to base your layered container on you can get that from your cluster like:
oc adm release info --image-for rhel-coreos
or from quay using a pull secret like:
oc adm release info --registry-config /path/to/pull-secret --image-for=rhel-coreos quay.io/openshift-release-dev/ocp-release:4.15.1-x86_64
where you can replace 4.15.1
with the version of OpenShift you are currently targeting.
Now you can do a container build. Here is an example Containerfile
that layers a single package from EPEL:
FROM scratch
#Enable EPEL (more info at https://docs.fedoraproject.org/en-US/epel/ ) and install htop
RUN rpm-ostree install https://dl.fedoraproject.org/pub/epel/epel-release-latest-9.noarch.rpm && \
rpm-ostree install podman-tui && \
ostree container commit
Note that in RHCOS 4.16 and newer, you can also use dnf install
instead of rpm-ostree install
.
And the command to build the container would look like:
RHCOS_CONTAINER='quay.io/openshift-release-dev/ocp-v4.0-art-dev@sha256:....'
PULL_SECRET=/path/to/pull-secret
podman build \
--from $RHCOS_CONTAINER \
--authfile $PULL_SECRET \
--file Containerfile \
--tag quay.io/myorg/myrepo:mytag
First, we need to convert the image to an OCI archive:
# to pull from local storage
skopeo copy containers-storage:quay.io/myorg/myrepo:mytag oci-archive:my-custom-rhcos.ociarchive
# OR to pull from a registry
skopeo copy --authfile /path/to/pull-secret docker://registry.com/org/repo:latest oci-archive:./my-custom-rhcos.ociarchive
You can now take that ociarchive and create a disk image for a
platform (i.e. qemu
, metal
or gcp
). First you need an
environment to run OSBuild in. Right now this needs to be a
fully up to date Fedora 40 machine with SELinux in permissive
mode and some software installed:
sudo dnf update -y
sudo setenforce 0
sudo sed -i -e 's/SELINUX=enforcing/SELINUX=permissive/' /etc/selinux/config
sudo dnf install -y --enablerepo=updates-testing osbuild osbuild-tools osbuild-ostree podman jq xfsprogs e2fsprogs
Now you should be able to generate an image with something like:
ociarchive=/path/to/my-custom-rhcos.ociarchive
platform=qemu
sudo ./custom-coreos-disk-images.sh $ociarchive $platform
Which will create the file my-custom-rhcos.ociarchive.x86_64.qcow2
in
the current working directory that can then be used.
You will also want to push
the custom container image to a registry and point OpenShift at it using a
MachineConfig with the osImageURL
field set to the image. Otherwise, upon
booting, the node will immediately be switched to the default OS image for
the target OpenShift version.
Create a MachineConfig like the following:
apiVersion: machineconfiguration.openshift.io/v1
kind: MachineConfig
metadata:
labels:
machineconfiguration.openshift.io/role: worker
name: custom-image
spec:
osImageURL: example.com/my/custom-image@sha256...
If scaling up, you can specify this MachineConfig as usual using oc apply -f
.
If installing a cluster, you can specify the MachineConfig at that point so that it's part of the initial bootstrapping. For examples of this, see the documentation at: