GCP Associate Cloud Engineer Practice Question 5: GKE Autopilot Cluster Mode
Google Cloud Associate Cloud Engineer · Test 1 · question 5 of 10
A platform team is moving a set of stateless microservices to GKE. They will keep managing their own Kubernetes manifests, Services, and HorizontalPodAutoscalers with kubectl, but they have no one to size, upgrade, patch, or scale worker nodes, and they want to be billed for the CPU and memory their Pods request rather than for provisioned node capacity.
Which approach meets these requirements?
Show answer and explanation
Answer:
Autopilot is the GKE mode of operation in which Google manages the node infrastructure: nodes are provisioned, scaled, upgraded, repaired, and hardened for you, and you never create or size node pools. For general-purpose workloads on the container-optimized compute platform you are billed for the CPU, memory, and ephemeral storage your Pods request rather than for whole nodes. Because it is still GKE, the team keeps the Kubernetes API, their manifests, and kubectl.
Option A leaves the team on Standard, where they still own node pools: machine type selection, node-pool upgrades, and capacity planning remain their job. Cluster autoscaler adds and removes nodes but does not change the node-based billing model or the operational burden.
Option C reduces the patching workload (auto-upgrade and auto-repair are genuinely useful), but it is still a Standard cluster billed per node, and it is zonal, so the control plane and nodes live in a single zone.
Option D would remove node management, but it discards the requirement to keep the existing Kubernetes manifests and objects. Cloud Run has no Kubernetes API, so Services and HorizontalPodAutoscalers would have to be rewritten.
Reference: Autopilot overview