You need to deploy an internet-facing microservices application to Google Kubernetes Engine (GKE). You want to validate new features using the A/B testing method. You have the following requirements…

Question

Options

• A
  1. Configure your CI/CD pipeline to update the Deployment manifest file by replacing the
• B
  1. Create a second namespace on GKE for the new release version.
• C
  1. Install the Anthos Service Mesh on your GKE cluster.
• D
  1. Implement a rolling update pattern by replacing the Pods gradually with the new release

• A
  11% (4)
• B
  6% (2)
• C
  81% (29)
• D
  3% (1)

Explanation

A/B testing requires running two versions simultaneously and splitting real production traffic between them by percentage or header-based rules - not just rolling one version out gradually. Anthos Service Mesh (based on Istio) provides a traffic management layer with VirtualService and DestinationRule resources that enable fine-grained, percentage-based routing between service versions. Both versions run concurrently, satisfying the no-downtime requirement, and a defined subset of real users is sent to the new release. Option A (replacing the Deployment manifest) performs a standard Deployment update - Kubernetes rolls the new image across pods but does not keep the old version alive for side-by-side comparison. Option B (a second namespace) is cumbersome and does not natively provide traffic splitting at the load-balancer level. Option D (rolling update) gradually replaces old pods with new ones, which is zero-downtime but is canary/rolling, not true A/B - both versions are not persistently available for controlled user cohorts.

No community discussion yet for this question.