GCP Professional Cloud Architect Practice Question

Your company runs a stateless Java API on Google Kubernetes Engine behind a ClusterIP Service. To release version 2.0, you update the manifest so that the Deployment uses the RollingUpdate strategy with maxSurge: 1 and maxUnavailable: 1. Site reliability engineers require that at least 9 of the existing 10 replicas remain available throughout the rollout and that clients are never switched to the new version all at once. Which outcome should the team expect when this update is applied?

All legacy Pods are terminated simultaneously; afterward ten version 2.0 Pods are created, resulting in a brief service outage.
The scheduler first creates ten new Pods running version 2.0, shifts all traffic to them at once, and then deletes the legacy Pods.
Traffic is initially routed to 5 % of the new version 2.0 Pods while 95 % continues to hit the legacy Pods; the percentages are gradually adjusted until 100 % is on version 2.0.
The Deployment alternates between creating one version 2.0 Pod and terminating one legacy Pod, guaranteeing a minimum of nine healthy Pods during the rollout.

Report Issue

Answer Description

With a Kubernetes Deployment that uses the RollingUpdate strategy, the controller gradually replaces Pods in the existing ReplicaSet with Pods from the new ReplicaSet. The parameters maxSurge: 1 and maxUnavailable: 1 mean the controller may create at most one additional Pod above the desired replica count and may have at most one Pod unavailable at any time. In practice, the scheduler creates one new version-2.0 Pod, waits until it is Ready, then terminates one legacy Pod, repeating the pattern until all Pods run the new version. This keeps at least 9 Pods (10 desired minus 1 allowed unavailable) serving traffic throughout the process and avoids an abrupt traffic switch. The other options describe blue/green, recreate, and canary behaviors, none of which match RollingUpdate with the specified limits.

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.