GCP Professional Cloud Architect Practice Question

A fintech team hosts a latency-sensitive REST API on Cloud Run, fronted by an external HTTP(S) load balancer that exposes api.example.com. For each release they must:

run functional and load tests against the new version without affecting customers,
switch all production traffic instantaneously after tests pass,
be able to roll back within one minute, and
avoid changing the public DNS name or load-balancer endpoint. Which deployment plan best implements a blue-green strategy while meeting every requirement?

Deploy the new version as another Cloud Run service in the same region, add it as a second backend to the HTTP(S) load balancer for testing, then edit backend weights to shift traffic when ready.
Create a new revision of the existing Cloud Run service, tag it for testing, then use Cloud Run traffic splitting to move production traffic from 0 % to 100 % to the new revision (and back if needed).
Deploy the new version as a separate Cloud Run service in a different project, verify it with tests, then update Cloud DNS to point api.example.com to the new project's load-balancer IP.
Replace the container image of the existing Cloud Run revision in place, using max-unavailable = 0 to ensure instances start before old ones terminate.

Report Issue

Answer Description

Blue-green on Cloud Run is easiest when both versions live as separate revisions of the same service. Deploying the container creates a new revision that is addressable by a tag for pre-production testing while the current revision continues to serve 100 % of user traffic. After tests pass, the traffic-splitting API (or console/CLI) can shift routing from 0 % to 100 % for the new revision in a single atomic operation; reversing the split provides an instant rollback. Because the service URL and the HTTP(S) load balancer backend remain unchanged, no DNS or load-balancer reconfiguration is required. Deploying to different services or projects would necessitate updating load-balancer backends or DNS, lengthening cut-over and rollback times and increasing risk. An in-place image update without a parallel green environment violates the blue-green principle of simultaneous production and candidate versions.

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.