GCP Compute Engine

Deploy, manage, and scale Compute Engine virtual machines on Google Cloud Platform.

When to Use

• Deploying web servers, application backends, or batch-processing workloads on GCP
• Running workloads that need full OS-level control (unlike Cloud Run or App Engine)
• Creating managed instance groups for auto-healing and auto-scaling behind a load balancer
• Provisioning GPU-attached VMs for ML training or rendering pipelines
• Cost-optimizing non-critical workloads with preemptible or spot VMs

Target, billable effects and recovery

Confirm project/account/zone, instance identity, costs and approved shutdown/resize scope. Review startup scripts as privileged code and pin trusted images; inspect service-account IAM rather than assuming cloud-platform scope is least privilege. Public IPs/firewall tags expose services only when matching rules exist; prefer required private access. The sample nginx startup does not define /healthz, so implement/test that route before attaching the shown MIG health check. Instance images/snapshots can contain secrets and are not application-consistent backups without appropriate quiescing. Shared-core table CPU values describe approximate baseline entitlement, not whole guest-visible vCPU topology.

Prerequisites