Exam Associate Cloud Engineer topic 1 question 255 discussion

B. Gradually roll out the new revision and split customer traffic between the revisions to allow rollback in case a problem occurs.

I would go with B. Here is why: Requirements: 1. Reduce the number of customers who might be affected by an outage (caused by a new version release) 2. Don't introduce any development or operational costs to the customers. (3. Follow Google-recommended practices) A: Might require the customer to setup exponential backoff on their end -> Requirement 2 not met. B: Fulfills all requirements. C: Does not meet requirement 1, as switching 100% traffic at once to the new version would affect all of the customers, if there would be any issue in this version. D: Does not meet requirements 1, 2 and 3, as an issue would affect all customers (once they have switched to the new service), switching to the second cloud run service causes dev/ops costs on the customer side and it doesn't follow Google-recommended practices, as a new version of an existing service should not be released as a new service.

B for the win! Deploy new features = split traffic. Every time.

Per ChatGPT, Option B aligns with the recommended practice of gradually rolling out new revisions, allowing for controlled monitoring and risk mitigation, without introducing any development or operational costs to customers.