Goal: 

As an administrator, I should be able to install a 3-node cluster (no workers) on GCP via the IPI or UPI installation paths and have a fully conformant OpenShift cluster (all features enabled).

Problem:

OpenShift is intended to run in a wide variety of environments. Over the years we have refined and reduced the default required footprint - first by removing the hard requirement for infrastructure nodes in 4.1, and second by preparing for 3-node clusters in 4.2 with the introduction of the masterSchedulable flag. OpenShift should continue to take steps to reduce the default footprint and incentivize improvements that allow it to work with smaller 3-node and eventually one node clusters.

Three-node compact cluster is already supported on baremetal, and we need to expand coverage to all other supported platforms.  The goal here is to ensure we have CI for this configuration, and identify any bugs/defects for flakes or faliure related to having a 3-node compact cluster on the public cloud, and open bugs so they can be actioned on.

Why is this important: 

• Enables OpenShift to be used for Telco edge use cases where compact clusters are required.
• Allows developers to run smaller clusters that are more resource efficient.

 
Lifecycle Information:

• Core

Dependencies (internal and external):

Prioritized epics + deliverables (in scope / not in scope):

• Allow pods to be schedulable on Master Nodes when a 3-node cluster is deployed
• Automatically place router pods on Masters if no worker nodes are present
• Ensure 3-node clusters work correctly in cloud environments for UPI and IPI deployment methods
• Support transitioning to a larger cluster and make the control plane unschedulable.
• Add compact 3-node cluster configurations to CI test system

Estimate (XS, S, M, L, XL, XXL): 

Previous Work:

Customers: 

Open questions: