Managed instance groups are the default answer for high availability on Compute Engine, but the Professional Cloud Architect exam will hand you scenarios where MIGs are exactly the wrong tool. When that happens, the architecture you reach for is the active-standby model. I want to walk through when MIGs break down and what the active-standby pattern actually looks like in practice.
There are three signals on the exam that tell you a MIG will not work for the workload in front of you.
The first is when horizontal scaling is not allowed but high availability is still required. MIGs are built around the idea that you can run multiple identical instances behind a load balancer and let traffic spread across them. Some applications cannot do that. They might rely on a single instance to maintain state, or they might require a configuration that cannot be duplicated across multiple VMs. If the application cannot scale out, a MIG cannot do its job.
The second signal is concurrent data access. MIGs assume stateless workloads where instances do not fight over shared storage. If your workload uses a persistent disk that only one instance can attach to at a time, multiple MIG instances will try to attach to the same disk and you get a conflict. Stateful workloads with strict control over data access do not belong in a MIG.
The third signal is when only one instance is allowed to handle traffic at a time. MIGs are designed to distribute load across instances. If the workload requires that exactly one instance receive traffic, a MIG fights against that requirement.
When you see any of these on the Professional Cloud Architect exam, your next thought should be active-standby.
The active-standby model gives you high availability without horizontal scaling. The structure is simple: two VMs in two different zones, a load balancer in front of them, and a regional persistent disk that can move between zones.
One VM is active and handles all incoming traffic. The other VM is in standby, idle but ready. The load balancer sends traffic to the active VM under normal conditions. If the active VM fails, the load balancer redirects traffic to the standby VM in the other zone.
The persistent disk is the piece that makes this work for stateful workloads. A regional persistent disk synchronously replicates across two zones in the same region, so it can be attached to instances in either zone. Under normal operation, the disk is attached to the active VM. When the active VM fails, the standby VM in the other zone attaches to the disk and continues serving. There is no concurrent access at any point. The disk is attached to exactly one VM at a time, which is what stateful workloads require.
The Professional Cloud Architect exam tests whether you can match a workload to the right Compute Engine pattern. MIGs are the right answer for stateless, horizontally scalable applications. Active-standby is the right answer when the workload cannot scale horizontally, cannot share data across instances, or cannot have more than one instance receiving traffic.
The pieces that signal active-standby in a question are: a load balancer with failover, two zones, a regional persistent disk, and a description of the workload that rules out horizontal scaling. When you see those together, the answer is active-standby, not a MIG.
My Professional Cloud Architect course covers the active-standby model alongside the rest of the compute material.
