GCP Associate Cloud Engineer Practice Question

Your team deploys a new microservice to an existing GKE Autopilot cluster. The Deployment YAML does not specify any CPU or memory requests or limits. After the rollout you run kubectl describe pod and notice each container now shows requests: cpu 250m, memory 512Mi, even though the developers expected to use far less. You must explain what happened and how they can request smaller values. Which statement correctly describes the situation?

The Horizontal Pod Autoscaler assigned initial resource requests automatically; defining an HPA with lower target utilization will allow the Pods to start with smaller requests.
Autopilot adds default requests of 0.25 vCPU and 512 MiB when none are specified; these defaults are also the minimum allowed, so requesting less is not possible.
Autopilot derives resource requests from the container image size; shrinking the image or using a scratch base will let the team achieve the 50 m/128 Mi requests they wanted.
Kubernetes reserves 10 % of each node for system daemons, so the scheduler rounded the developers' numbers up; creating a custom RuntimeClass is needed to override the reservation.

Report Issue

Answer Description

In Autopilot clusters Google manages the underlying nodes and enforces minimum resource requests so that every Pod is guaranteed a baseline level of CPU and memory. If a container spec omits the resources field, Autopilot injects default requests that currently equal the minimum allowed-0.25 vCPU (250 m) and 512 MiB. Any value below these minima or outside the valid CPU-to-memory ratio is automatically adjusted upward. Because those defaults represent the lowest permitted values, the team cannot request less; instead they must explicitly set requests that are equal to or greater than the enforced minimums if they want different numbers. The other options are incorrect: the values are not the result of Kubernetes node-level reservations, horizontal pod autoscaling, or container image size.

Ask Bash

Bash is our AI bot, trained to help you pass your exam. AI Generated Content may display inaccurate information, always double-check anything important.