Architecture Deep Dive

Scale-to-Zero: One Pod, On Demand

How a dev or QA engineer hitting an ephemeral test URL triggers exactly one Pod to spin up for them — and how it scales back down to nothing the moment they're done. Drag the scene to look around, then click the glowing ring to fire a test request.

Dev / QAOpenShift RouteKnative Activator + KPAPod bot — sleeping → drowsy → awake

Azure Red Hat OpenShift

OpenShift Serverless (built on Knative Serving) is the native way to get scale-to-zero on ARO — no extra controllers to self-host.

A dev/QA engineer opens the ephemeral test URL. The OpenShift Route points at a Knative Service that currently has zero running Pods.
The Knative Activator sits in the data path while the Service is scaled to zero. It buffers the request instead of returning a 502.
The Activator reports the pending request to the Knative Pod Autoscaler (KPA), which scales the Deployment from 0 → 1 replica.
Once the new Pod passes its readiness probe, the Activator forwards the buffered request straight to it and steps back out of the path.
After the configured idle window (default ~30s) with no further requests, the KPA scales the Deployment back down to zero.