GCP Professional Cloud Architect Practice Question

Your e-commerce team runs a Go web API on Cloud Run in us-central1. A recent regional outage caused a 3-hour downtime. They want to update their Cloud Build pipeline to deploy the code (no Dockerfile) to an additional europe-west1 service, release each new revision to only 10 % of users first, and achieve 99.95 % global availability while avoiding inter-region egress for session writes stored in Firestore. Which deployment architecture best meets these goals?

Deploy a single Cloud Run service with the --region=global flag so Cloud Run automatically spans multiple regions; use its built-in traffic splitting for the 10 % canary while keeping a single multi-region Firestore database.
Create separate Cloud Run services in us-central1 and europe-west1. Use Cloud Build to run gcloud run deploy --source for each region, enable 10 / 90 traffic splitting between revisions on every service, and place both services behind an external HTTP(S) Load Balancer with serverless NEGs configured for geo-based routing; point each service to its local regional Firestore instance.
Migrate the workload to two GKE Autopilot clusters (us-central1 and europe-west1) and expose them through multi-cluster Ingress; handle canary releases with Kubernetes rolling updates and continue using the existing Firestore instance in us-central1.
Refactor the service into Cloud Functions deployed in us-central1 and europe-west1, front them with Cloud CDN, and enable gradual rollout by versioning the function; store session data in a single multi-region Firestore database to simplify consistency.

Report Issue

Answer Description

Deploying two independent Cloud Run services-one in us-central1 and one in europe-west1-lets the application survive a regional failure and meets the 99.95 % availability target when traffic is routed through a global external HTTP(S) Load Balancer backed by serverless NEGs. Geo-based routing sends users to the closest region, so the Firestore traffic stays within the same continent and avoids cross-region egress. Using gcloud run deploy --source in the Cloud Build pipeline compiles the container image from source without a Dockerfile. Cloud Run traffic-splitting on each service can direct 10 % of traffic to the new revision for a canary and provides an instant rollback path. The other options either rely on a single regional service, use an unsupported "multi-region" Cloud Run setting, or migrate to a different compute platform, so they do not simultaneously satisfy all stated requirements.

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.