Prepare and test your skills

Prepare and test your skills

Free sample question with answers and detailed explanation below. The Professional Cloud Developer practice bank has an additional 608 practice questions.Try 10 free realistic questions: instant score, no credit card, and no account required.
Your team manages a stateless web API on Google Kubernetes Engine (GKE) that experiences sudden traffic spikes of up to 30% within two to three minutes during promotional events. Recent monitoring shows two recurring issues:
You need to establish accurate per-Pod resource baselines and configure autoscaling policies that prevent OOM terminations and absorb sudden load bursts without over-provisioning during quiet periods.
What should you do?
This approach bypasses application profiling in favor of high static Pod requests, scales GKE node pools down to zero nodes when idle, and relies on Compute Engine predictive autoscaling to anticipate short promotional traffic bursts.
You were managing a non-Kubernetes Compute Engine Managed Instance Group (MIG) with long application initialization times and predictable daily or weekly CPU cycles lasting longer than 10 minutes.
This configuration attempts to automate both vertical and horizontal scaling simultaneously on CPU, caps CPU usage strictly at the request value, leaves memory uncapped, and delays horizontal scaling until Pods reach 95% CPU utilization.
kubelet to evict or OOM-kill Pods and destabilizing the node.There is no scenario where combining unbounded memory limits with a 95% HPA CPU target is optimal for a bursty serving API. However, Multidimensional Pod Autoscaling (MPA) can safely combine horizontal CPU scaling with vertical memory scaling when configured properly.
This approach combines controlled load benchmarking, Vertical Pod Autoscaler (VPA) profiling, Kubernetes Quality of Service (QoS) best practices, and a buffered Horizontal Pod Autoscaler (HPA) target to establish accurate baselines and maintain responsive scaling.
(1 - buffer) / (1 + expected_growth)—for example, (1 - 0.1) / (1 + 0.3) = 0.69 (69% CPU)—ensures HPA triggers scale-out early enough to absorb a 30% burst over 2–3 minutes before Pods saturate at 100% CPU.Unlike relying on static guesses or high utilization targets, combining empirical single-Pod load testing with VPA recommendation mode produces data-driven baselines, while a mathematically buffered HPA threshold accounts for container startup and node provisioning latency.
This configuration uses historical 30-day average metrics to size Pods, triggers HPA scaling at 90% memory utilization, runs serving APIs on Spot VMs, and waits for Pods to become Pending before provisioning nodes.
Pending before Cluster Autoscaler provisions new nodes adds minutes of infrastructure startup latency.You were running fault-tolerant, memory-steady batch processing jobs on Spot VMs where startup latency is acceptable and cost reduction is the sole priority.
Answer them yourself and get your score the moment you finish. No account needed.
Full access includes
Keep the momentum going with these hand-picked practice scenarios
Last updated Validated