GCP Professional Cloud Architect Practice Question

Your team must deploy a stateless Java API on Compute Engine and meet the following production goals:

Run the service in three zones of us-central1.
Keep between 3 and 30 VMs online, scaling on average CPU load.
Ensure at least one healthy VM remains in every zone during image rollouts.
Automatically recreate any VM whose OS or application health check fails. Which deployment design satisfies all of these requirements while minimizing operational effort?

Create one regional managed instance group with a minimum size of 3 and a CPU-based autoscaler capped at 30, set the distribution shape to EVEN, attach an HTTP health check for autohealing, and use a rolling-update policy of maxUnavailable = 1 and maxSurge = 0.
Run a single zonal managed instance group with a minimum of 3 instances, enable host live migration for maintenance, and distribute traffic across zones by attaching the group to a global load balancer.
Deploy three separate zonal managed instance groups, each with a minimum of 1 instance and a CPU autoscaler up to 10; place them behind an external HTTP(S) load balancer and perform full group restarts when updating the instance template.
Use a regional managed instance group composed entirely of preemptible VMs, scale between 3 and 30 instances, disable autohealing, and enable proactive rolling updates that replace 25 % of instances simultaneously.

Report Issue

Answer Description

A regional managed instance group (MIG) automatically spreads VMs across the zones in its region. Setting the distribution shape to EVEN and a minimum size of 3 guarantees one VM per zone at all times. An autoscaler with a maximum size of 30 handles CPU-based scaling. Attaching an HTTP or TCP health check enables autohealing so failed instances are recreated. During rolling updates, configuring maxUnavailable to 1 and maxSurge to 0 lets the group replace one instance at a time without creating extra VMs, so each zone always keeps at least one healthy instance. The other options miss one or more goals: separate zonal MIGs require manual coordination to preserve per-zone capacity during updates, a single zonal MIG cannot survive a zonal outage, and using preemptible VMs without autohealing does not meet availability or health-check requirements.

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.