GCP Professional Cloud Architect Practice Question

A game studio runs its real-time leaderboard on two GKE Autopilot clusters behind a global external HTTP(S) load balancer. Normal peak is 50 000 concurrent players, but an upcoming tournament may triple traffic. The SRE team wants to run a one-time, two-hour stress test that drives about 300 000 concurrent requests, originates traffic from several Google Cloud regions, streams latency and error metrics to Cloud Monitoring during the run, and removes all load-generator resources when the test finishes. Which approach best satisfies these requirements?

Create Cloud Functions in several regions that continuously invoke the leaderboard endpoint; examine logs afterward and delete the functions manually.
Schedule a Google Cloud Armor red-team DDoS simulation against the load balancer and rely solely on the post-exercise report for performance insights.
Use the Distributed Load Testing on Google Cloud solution: create lightweight Autopilot GKE clusters in multiple regions with Terraform, deploy Locust master and worker pods to those existing clusters, stream metrics to Cloud Monitoring, and run terraform destroy afterward to tear everything down.
Deploy the Cloud Spanner benchmarking tool on a regional managed instance group in Compute Engine and review results later in Cloud Trace.

Report Issue

Answer Description

The Distributed Load Testing on Google Cloud solution runs Locust master and worker pods on GKE. You can use Terraform to spin up lightweight Autopilot clusters in the desired regions, deploy the solution to each cluster, and start the test. Locust's distributed design lets each regional cluster generate hundreds of thousands of concurrent HTTP requests. The solution exports latency, throughput, and error metrics to Cloud Monitoring in near real time, where the SRE team can build dashboards. After the two-hour run, the same Terraform configuration can be used with terraform destroy to delete the clusters and any ancillary resources, eliminating ongoing cost. The other approaches either rely on tools that are not designed for HTTP load testing, generate traffic from only one region, lack real-time Cloud Monitoring visibility, or require manual cleanup, so they do not meet all stated goals.

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.