GCP Professional Cloud Architect Practice Question

Your team operates a stateless web application on a regional Standard GKE cluster. The Deployment currently runs 10 replicas. A new container image must be released during peak hours without reducing serving capacity below the existing 10 replicas at any point in time. The team is willing to tolerate up to 20 % temporary over-provisioning to achieve this and wants to minimize excess node usage. Which Deployment update strategy best meets the requirement?

RollingUpdate strategy with both maxSurge and maxUnavailable set to 20 %
Recreate update strategy (all existing Pods are terminated before new Pods start)
RollingUpdate strategy with maxSurge set to 0 and maxUnavailable set to 1
RollingUpdate strategy with maxSurge set to 2 and maxUnavailable set to 0

Report Issue

Answer Description

To guarantee that all 10 replicas remain available throughout the rollout, the Deployment must never mark a running Pod unavailable while the replacement Pod is still starting. Setting the strategy to RollingUpdate with maxUnavailable = 0 prevents any replica from being taken out of service before its replacement is ready. Limiting maxSurge to 2 (20 % of 10) respects the over-provisioning ceiling while allowing two new Pods to start at a time, so the total replica count will vary between 10 and 12 during the update and never drop below 10.

The Recreate strategy stops every Pod before starting the new ones, causing complete downtime. A RollingUpdate with maxUnavailable = 1 or with both maxSurge and maxUnavailable at 20 % allows the replica count to fall to 9 or 8 respectively, violating the zero-capacity-loss requirement. A configuration with maxSurge = 0 and maxUnavailable = 0 cannot progress because no new Pods can start and no old Pods can be terminated.

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.