Document URL: https://docs.openshift.com/container-platform/4.8/machine_management/adding-rhel-compute.html

Section Number and Name:

Describe the issue: Documentation is confusing around the addition of RHEL worker nodes in that it seems to indicate UPI and IPI is supported but then only shows UPI references further in the documentation.

Suggestions for improvement: Can we provide further details as to why the UPI process in this use case applies to both UPI and IPI deployed environments when adding RHEL worker nodes

Additional information:

Openshift 4 allows you to run RHEL 7 as your worker nodes (masters must be RHCOS). The MCD does run on RHEL nodes, and does a weird half-management scheme of said node, in which it updates the files/units etc., but does not update the OS or packages (such as kubelet). This means during an Openshift upgrade the cluster auto-upgrades half, and then you have to do the other half via ansible scripts.

Provisioning during firstboot is also handled by ansible. Since RHEL doesn’t run ignition, ansible instead runs the MCD in “onceFrom” mode to perform a write of the system as if you had attempted to update from “none” to “rendered worker config” with a real MCD pod.