Skip to main content

Google Compute Engine

Google Compute Engine (GCE) is an Infrastructure as a Service (IaaS) component of Google Cloud Platform that allows users to create and manage virtual machines (VMs) on demand. It provides scalable, high-performance VMs that run on Google’s infrastructure.

1. Enabling GCE API​

Before creating a GCE instance, the Compute Engine API must be enabled. This can be done manually or via Infrastructure as Code (IaC) tools. Enabling the API is a prerequisite for provisioning and managing VM instances.

2. Creating a Virtual Machine Instance​

When configuring a new GCE VM instance, several parameters must be specified:

A. Machine Configuration​

  • Machine Type: Choose from predefined (e2, n1, n2, etc.) or custom machine types with specific vCPU and RAM.

  • OS Image and Boot Disk:

    • Choose OS (Debian, Ubuntu, Red Hat, Windows, etc.)
    • Boot disk type: Standard persistent disk (PD), SSD PD, or balanced PD.
    • Optionally deploy container images (e.g., Docker images) directly.

B. Networking Configuration​

  • Allow HTTP/HTTPS Traffic: Enable firewalls to permit incoming web traffic.
  • Load Balancer Health Check: Required for backends in load balancing.
  • IP Forwarding: Enable if the VM will act as a NAT gateway or VPN endpoint.
  • Network Bandwidth: Optional; controls egress throughput.
  • Network Interfaces: Define subnet, IP ranges, and tags for routing/firewall rules.

C. Observability (Monitoring and Logging)​

  • Install Ops Agent: For system metrics and logs ingestion into Cloud Monitoring and Cloud Logging.
  • Enable Display Device: For GUI-based interaction (mostly for Windows VMs).

D. Security Settings​

  • Access Scopes: Define access level for default service account (default/full/custom per API).
  • Shielded VM Options: Protect against rootkit and bootkit attacks (secure boot, vTPM, integrity monitoring).

E. Advanced Options​

  • Deletion Protection: Prevent accidental deletions.
  • Reservations: Use capacity reservations for guaranteed availability.
  • Startup Scripts: Execute commands at boot time (e.g., software install).
  • Metadata: Pass custom metadata or scripts.
  • Encryption: Use Google-managed keys or customer-supplied keys (CSEK/CMEK).
  • Provisioning Model: Choose between standard, spot, or preemptible instances.

F. Sole Tenancy​

  • Host your VM on dedicated physical machines, useful for licensing or isolation purposes.

3. “Equivalent Code” Option​

This generates the corresponding gcloud CLI command or Terraform code for the selected configuration, facilitating automation or reproducibility.

4. Static and Dynamic IP Addressing​

Internal IP:​

  • Always assigned, used for intra-VPC communication.

External IP:​

  • Ephemeral IP: Assigned temporarily; changes on VM restart.
  • Static IP: Reserved and fixed; useful for DNS mapping or persistent public access.

Reserving a Static IP:​

  • Select network service tier (Standard or Premium).
  • Choose IP version (IPv4 or IPv6).
  • Decide on scope: Regional (for GCE) or Global (for load balancers).
  • Attach it to the VM from the reserved list.
  • Note: Static IPs are billed even when not in use.

5. VM Creation Methods​

  1. Startup Script:

    • Run shell or PowerShell scripts at boot time.

  2. Instance Template:

    • Predefined configuration template to deploy uniform VMs.
    • Immutable after creation (clone and modify to change).

  3. Custom Image:

    • Snapshot of an existing configured VM.
    • Useful for replicating pre-installed software environments.

6. Sole-Tenant Nodes​

A Sole-Tenant Node is a physical machine dedicated to a single tenant. Recommended for:

  • Compliance requirements
  • Workloads requiring licensing isolation

Steps to Configure:​

  • Create Node Group: Define region, zone, and node group name.

  • Create Node Template:

    • Node type, SSD configuration, GPU acceleration (e.g., NVIDIA T4)
    • CPU overcommit options
    • Affinity labels for scheduling control

  • Auto Scaling: Configure scale-out or fixed number of nodes.

  • Maintenance Policy: Decide live migration or VM restart behavior.

  • Sharing Policy: Share nodes across projects or services if needed.

7. VM Manager​

Useful for managing large-scale VM fleets:

  • OS Patch Management: Schedule and enforce OS patches across VMs.
  • OS Configuration Management: Apply consistent settings like SSH access, packages, and services.

8. Instance Groups​

GCP provides two types:

A. Managed Instance Group (MIG)​

  • Identical instances created from an instance template.

  • Supports:

    • Auto-healing (health check based)
    • Auto-scaling (based on load)
    • Rolling updates and version control
    • Canary deployments

MIG Creation Requires:​

  • Instance template

  • Auto-scaling policy:

    • Metrics (CPU, Load Balancer, etc.)
    • Cooldown period
    • Min/Max instance limits
    • Scale-in control

  • Health check for auto-healing

B. Unmanaged Instance Group​

  • Heterogeneous VM instances.
  • No auto-scaling or auto-healing.
  • Deprecated for production workloads; useful for grouping unrelated VMs.

Additional Notes​

  • Project Settings: Every resource is part of a GCP project.
  • Billing: Each project must be linked to a billing account.
  • IAM Permissions: Role-based access control is critical for team environments.
  • API Enablement: Compute Engine API must be enabled for all operations.