Managing and provisioning a solution infrastructure · Topic 4 of 6
Configuring Compute Systems
Configuring managed instance groups, Spot and standard VMs, GKE clusters and upgrades, Cloud Run, serverless networking, VMware Engine and patch management.
- 4 min
- Instance templates plus regional MIGs for resilient, autoscaled VMs.
- Canary rolling updates need little extra capacity; blue-green needs a second full set.
- Spot for interruptible work only.
What it is
Configuring compute means provisioning the chosen platform so it scales, stays patched, connects to the right networks and costs no more than it should. The exam tests the settings behind reliable rollouts and efficient capacity.
How it works
Compute Engine provisioning
| Feature | Purpose |
|---|---|
| Instance templates | Define machine type, image, disks, service account and network once |
| Managed instance groups (MIGs) | Identical VMs with autoscaling, autohealing and rolling updates; regional MIGs spread across zones |
| Rolling and canary updates | Apply a new template to a percentage of instances first, with max surge and max unavailable settings |
| Custom images and image families | Golden images that always point to the latest approved build |
A canary update on a MIG updates a small share in place, which suits a project with little spare quota. Blue-green with a second full MIG doubles the VM count during the change.
Spot versus standard VMs
| Standard | Spot | |
|---|---|---|
| Price | Full, discounts through commitments | Much lower |
| Interruption | Only for maintenance (live migration where supported) | Can be stopped at any time when capacity is needed |
| Fits | Production services, stateful systems | Batch, CI, rendering, fault-tolerant workers |
Mix them: a standard baseline with Spot capacity for burst work, or Spot node pools in GKE for interruptible Pods.
GKE configuration
| Setting | Why it matters |
|---|---|
| Regional cluster | Control plane and nodes across zones, survives a zone loss |
| Autopilot or Standard | Who manages nodes and how billing works |
| Cluster autoscaler and node auto-provisioning | Add nodes and node pools as Pods need them |
| Horizontal and vertical Pod autoscaling | Scale replicas or Pod requests |
| Release channels and maintenance windows | Control when upgrades happen |
| Surge or blue-green node pool upgrades | Surge replaces nodes gradually; blue-green keeps old nodes for a soak period so you can roll back |
| Image streaming with Artifact Registry | Large images start before they finish downloading |
Serverless computing and networking
- Cloud Run scales on request concurrency; set minimum instances to avoid cold starts and maximum instances to protect downstream systems.
- Reach VPC resources from Cloud Run with Direct VPC egress or a Serverless VPC Access connector.
- Control who can call a service with IAM, ingress settings and, for browsers, IAP or a load balancer.
- Revisions and traffic splitting support canary releases and instant rollback; tags give a revision its own URL with no traffic.
Google Cloud VMware Engine
VMware Engine runs vSphere private clouds on dedicated nodes in Google Cloud. It joins your VPC through private connectivity, so migrated VMs reach native services such as BigQuery and Cloud Storage. Use it to exit a data center quickly without converting VMs.
Infrastructure orchestration and patch management
| Need | Tool |
|---|---|
| Declarative provisioning | Terraform, run by Infrastructure Manager or your pipeline |
| OS patching and inventory for VMs | VM Manager (OS patch management, OS inventory, OS policies) |
| Keeping containers patched | Rebuild images regularly, scan in Artifact Registry, use GKE release channels |
When the stem says nobody should patch servers, prefer serverless or Autopilot; when VMs must stay, VM Manager automates the patching.
How the exam asks about this
- "Rollback quickly after a node pool upgrade, extra nodes acceptable." Blue-green node pool upgrade with a soak time.
- "New image to 10 percent of 40 VMs first, little spare quota." A canary rolling update on the MIG.
- "Pods take four minutes to start because of a 7 GB image." Artifact Registry with image streaming.
What to remember for the exam
- Instance templates plus regional MIGs for resilient, autoscaled VMs.
- Canary rolling updates need little extra capacity; blue-green needs a second full set.
- Spot for interruptible work only.
- GKE: regional clusters, autoscaling, release channels, blue-green upgrades for safe rollback.
- Cloud Run: minimum instances for spikes, Direct VPC egress for private resources, tags for testing revisions.
A surge upgrade replaces old nodes as it goes, so there is nothing left to roll back to once it finishes.
Common mistakes
Putting a stateful service only on Spot VMs. It can be stopped at any time.
Capping Cloud Run maximum instances at today's peak for a tenfold surge.
Patching VMs by hand when VM Manager can schedule it.
Check yourself
Which GKE upgrade strategy keeps the old nodes for a soak period?
Blue-green node pool upgrade.
How does a Cloud Run service reach a private Cloud SQL IP in the VPC?
Through Direct VPC egress or a Serverless VPC Access connector.
Which tool schedules OS patching across Compute Engine VMs?
VM Manager OS patch management.
Topic 4 of 6 in Managing and provisioning a solution infrastructure
Test yourself on Managing and provisioning a solution infrastructure
The full Google Cloud Professional Cloud Architect question bank has 877 questions, with timed practice exams and an explanation for every answer.
This study guide is independent and is not affiliated with, endorsed by, or sponsored by Google Cloud. Vendor names, certification names and exam codes are trademarks of their respective owners and are used for identification only.