GCP Professional Cloud Architect Practice Question

Your e-commerce platform runs microservices in two Google Cloud projects (prod-us and prod-eu) and an on-premises Kubernetes cluster that already exports Prometheus metrics. You need a single dashboard showing aggregated request latency, error rate, and custom KPIs for the checkout service across all environments, and an alert when the checkout service's p99 latency exceeds 300 ms for 5 continuous minutes. Which solution provides these capabilities with the least operational overhead?

Use Cloud Logging exclusively: create logs-based metrics for latency, route all logs to Pub/Sub, and deploy a Cloud Function subscriber that aggregates the logs across projects and on-prem to generate custom alerts when p99 latency passes 300 ms.
Deploy independent Prometheus instances in each environment and federate them to a central Prometheus server on Compute Engine; build Grafana dashboards and use Alertmanager to notify when p99 latency is above 300 ms for five minutes.
Export metrics from Cloud Monitoring in each project to BigQuery, construct a Data Studio report that joins the datasets, and run a scheduled Cloud Function every five minutes to query BigQuery and send alerts if checkout latency exceeds 300 ms.
Create a single Cloud Monitoring Workspace in a new operations project, add the prod-us and prod-eu projects, deploy the Cloud Monitoring/Managed Service for Prometheus collector to the on-prem cluster, define a checkout service SLO with a 300 ms p99 latency objective, and configure an alerting policy that fires after five minutes of SLO violation.

Report Issue

Answer Description

Cloud Monitoring Workspaces can ingest metrics from multiple Google Cloud projects as well as external sources. Adding both projects to one workspace yields a unified dashboard without extra aggregation code. The Managed Service for Prometheus (or Prometheus receiver in the Ops Agent) lets you forward on-prem Prometheus metrics directly into Cloud Monitoring so they appear alongside GCP metrics. Within the workspace you can define a custom service, attach an SLO whose objective is p99 latency ≤ 300 ms, and create an alerting policy that triggers after five minutes of burn-rate violation-meeting the requirement without running or maintaining additional infrastructure.

The other approaches require you to operate and scale Prometheus, Grafana, Alertmanager, BigQuery jobs, or custom Cloud Functions, increasing operational burden and complexity while duplicating functionality already built into Cloud Monitoring.

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.