GCP Professional Cloud Architect Practice Question

Your team operates a containerized payment API on Cloud Run, exposed through a global external HTTPS load balancer and protected by Cloud Armor. A major code refactor (v2) must be released with zero production downtime and an instant rollback path if unexpected errors appear. You also need to run smoke tests against v2 before any user traffic reaches it. Which deployment approach best satisfies these blue-green requirements on Cloud Run?

Create a separate Cloud Run service for v2 in a new GCP project, complete testing there, then update public DNS to point to the new service once it is validated.
Deploy v2 as a new tagged revision in the same Cloud Run service, keep its traffic allocation at 0 %, run smoke tests via the tag URL, then shift traffic to 100 % when verified and back to the prior revision if needed.
Replace the existing revision in place with v2 and rely on Cloud Run's instance concurrency controls to prevent downtime; roll back by redeploying v1 if problems occur.
Deploy v2 as a new revision and immediately split production traffic 50 / 50 between v1 and v2, monitoring metrics before gradually increasing v2's share.

Report Issue

Answer Description

Creating a second revision of the existing Cloud Run service and tagging it as the "green" environment lets the platform start the new containers without sending them user traffic. Because traffic is initially set to 0 %, you can reach the revision directly using its tag URL to run validation tests. When v2 is approved, updating the service's traffic split to 100 % green / 0 % blue switches production instantly with no DNS or load-balancer changes. If issues arise, a single traffic-split command reverts back to the previous revision just as quickly. Updating a revision in place or relying on rolling updates does not provide an untouched "blue" fallback, while deploying to a separate project or distributing 50 % traffic before testing both risk user impact.

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.