Prerequisites

• OpenShift cluster with existing layered MachineConfigPool
• MachineOSConfig already configured and image built
• Layered image available in registry
• Understanding of Machine Config Daemon (MCD) behavior

Problem Statement

• In image mode, when you customize a node OS (via MOSC -> MOSB -> new image), nodes must reboot to pick up the new image.
• Today, this can lead to redundant reboots, including:
  • One reboot for base OS updates
  • One reboot for extensions/packages
  • One reboot for when a cluster joins a layered pool
  • Possibly another reboot during cluster upgrades

• Reboots are disruptive, so the ask is to make fewer and more coordinated reboots.

Specifically, we want to target the reboots associated with adding a new node to a layered pool, which require 2 OpenShift-managed reboots:

1. Initial Cluster Join: Node joins as worker, gets base configuration, reboots
2. Pool Assignment: Node is assigned to layered pool, MCD applies layered image, reboots again

Goal: Achieve a single reboot by having the node join the cluster already assigned to the layered pool, eliminating the intermediate worker pool membership.

Note: In addition to the 2 OpenShift-managed reboots, the node boots a third time when it initially starts up from the disk image. So total lifecycle boots is 3 from when a node starts up to joining a layered pool.