GCP Professional Cloud Architect Practice Question

Your retail analytics platform runs entirely in a single-region GKE cluster located in europe-west1. The PostgreSQL database is hosted on Cloud SQL. Compliance now mandates a disaster-recovery posture that guarantees a recovery point objective (RPO) of 15 minutes and a recovery time objective (RTO) of 60 minutes if the whole europe-west1 region is lost. Budget owners do not want to pay for a fully provisioned duplicate production environment running at all times. Which approach best meets the new objectives while keeping ongoing costs low?

Enable regional persistent disks and multi-zonal node pools within europe-west1 and rely on the region's three zones to stay available during failures.
Add a cross-region asynchronous Cloud SQL replica in europe-west4 and store GKE manifests in Terraform; on disaster, promote the replica and redeploy the cluster with Cloud Build/Deploy (warm standby).
Run full-capacity GKE and Cloud SQL deployments in both europe-west1 and europe-west4 behind an external HTTP(S) load balancer, using bidirectional replication for an active/active setup (hot standby).
Configure daily automated backups of Cloud SQL to a multi-region Cloud Storage bucket and recreate the GKE cluster and database from the latest backup if a disaster occurs (cold standby).

Report Issue

Answer Description

A warm-standby pattern strikes the desired balance between cost and recovery targets. A cross-region asynchronous Cloud SQL replica in europe-west4 normally keeps replication lag to seconds, easily meeting a 15-minute RPO. Because the replica is already provisioned, it can be promoted to a primary instance quickly; the application layer can then be redeployed from existing Terraform configurations and pipelines, bringing the service back within the 60-minute RTO.

Cold-backup strategies that rely solely on daily exports or snapshots would require rebuilding both the database and the GKE cluster, making it very difficult to restore service within one hour. A fully active/active multi-region deployment would meet (and exceed) the objectives but violates the cost constraint by running duplicate, full-sized environments continuously. Relying only on multi-zonal resources inside europe-west1 does not address a complete regional outage and therefore fails both RPO and RTO 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.