Skip to content
All posts

ROSA, ARO, and self-managed: which OpenShift cores are billable

· 6 min read

“OpenShift” is not one license. It is a family of offerings, and each one decides which cores are billable in a different way. Reconcile a cluster under the wrong model and the gap is wrong from the start — sometimes over-stated, sometimes under-stated, always indefensible at the negotiating table.

The offerings, at a glance

  • Self-managed OpenShift — you run the platform on your own infrastructure and hold a subscription sized in core-pairs.
  • ROSA (Red Hat OpenShift Service on AWS) — a managed service, metered and billed through the cloud provider.
  • ARO (Azure Red Hat OpenShift) — the Azure-native managed equivalent, with its own consumption model.
  • OpenShift Dedicated — a Red Hat-managed cluster on cloud infrastructure, subscribed by capacity.

Why the model decides the number

On self-managed OpenShift, the unit of entitlement is the core-pair, and your obligation is the core-pairs your billable nodes consume — resolved per SKU and per node role. On the managed offerings, the cluster is metered by the provider, so the “count” you reconcile is a consumption figure, not a fixed subscription of core-pairs.

Treat a managed cluster as if it were self-managed — or vice versa — and you compare the wrong entitlement against the wrong usage. The error is easy to make when one organization runs several offerings at once, which is increasingly the norm.

Where it goes wrong in practice

  • Mixed estates. A self-managed cluster on-prem and a ROSA cluster in AWS are two different licensing conversations counted in one spreadsheet.
  • Role assumptions. Which nodes are billable depends on the offering; counting control-plane and worker nodes the same way distorts the position.
  • SKU drift. A single environment can map to several Red Hat SKUs; if the SKU is mis-resolved, the entitlement is wrong before any math runs.

Getting it right per environment

The position is knowable for every offering — it just has to be measured with that offering’s own rules. That is what RenewalIntel does for Red Hat: it resolves the SKU, identifies the offering, applies the right entitlement model, and prices the gap with the calculation shown — so a mixed OpenShift estate reconciles to one defensible number instead of a guess.