examples

GKE Terraform Modules Example

This example demonstrates how to use the GKE cluster and node pools Terraform modules to create a complete Google Kubernetes Engine setup.

What This Example Creates

• A GKE cluster with private nodes
• Three different node pools:
  • Default pool: General-purpose workloads with autoscaling (1-3 nodes)
  • High-memory pool: Memory-intensive workloads with taints (0-2 nodes)
  • Spot pool: Cost-effective batch workloads using spot instances (0-5 nodes)

Prerequisites

1. GCP Project: You need a GCP project with the following APIs enabled:

   • Kubernetes Engine API
   • Compute Engine API
   • IAM API

2. Authentication: Configure authentication using one of:

   • gcloud auth application-default login
   • Service account key file
   • Workload Identity (for GKE)

3. Terraform: Install Terraform >= 1.0

Usage

1. Copy the example configuration:

   cp terraform.tfvars.example terraform.tfvars

2. Edit terraform.tfvars:

   • Replace your-gcp-project-id with your actual GCP project ID
   • Adjust other values as needed (region, cluster name, etc.)

3. Initialize Terraform:

   terraform init

4. Plan the deployment:

   terraform plan

5. Apply the configuration:

   terraform apply

6. Connect to your cluster:

   gcloud container clusters get-credentials my-gke-cluster --location us-central1 --project your-gcp-project-id

Configuration Options

Network Configuration

By default, this example uses the default VPC network. For production use, consider:

• Creating a custom VPC network
• Using VPC-native networking with secondary IP ranges
• Configuring firewall rules appropriately

Private Cluster

The example creates a private cluster with:

• Private nodes (no external IP addresses)
• Public endpoint (can be changed to private)
• Master authorized networks for access control

Node Pools

The example includes three node pools with different characteristics:

1. Default Pool (e2-medium):

   • General workloads
   • Auto-scaling 1-3 nodes
   • Standard persistent disk

2. High-Memory Pool (e2-highmem-2):

   • Memory-intensive workloads
   • Auto-scaling 0-2 nodes
   • SSD persistent disk
   • Tainted for specific workloads

3. Spot Pool (e2-standard-2):

   • Batch/fault-tolerant workloads
   • Auto-scaling 0-5 nodes
   • Spot instances for cost savings
   • Tainted for spot workloads

Customization

Adding More Node Pools

You can add additional node pools by extending the node_pools variable:

node_pools = {
  # ... existing pools ...
  
  "gpu-pool" = {
    machine_type = "n1-standard-2"
    disk_size_gb = 100
    gpu_config = {
      type  = "nvidia-tesla-t4"
      count = 1
    }
    autoscaling = {
      min_node_count = 0
      max_node_count = 2
    }
    taints = [
      {
        key    = "nvidia.com/gpu"
        value  = "true"
        effect = "NO_SCHEDULE"
      }
    ]
  }
}

Modifying Cluster Settings

You can customize various cluster settings:

# Enable different features
enable_workload_identity = true
network_policy_enabled   = true
release_channel         = "RAPID"  # or "REGULAR", "STABLE"

# Configure maintenance window
maintenance_start_time = "02:00"  # 2:00 AM

# Add resource labels
resource_labels = {
  environment = "production"
  team        = "platform"
  cost_center = "engineering"
}

Outputs

After successful deployment, you'll get outputs including:

• Cluster name and endpoint
• Node pool information
• kubectl connection command
• Network configuration details

Cleanup

To destroy all resources:

terraform destroy

Security Considerations

1. Master Authorized Networks: Configure appropriate CIDR blocks
2. Private Nodes: Use private nodes for better security
3. Workload Identity: Enable for secure pod-to-GCP-service authentication
4. Network Policies: Enable for pod-to-pod traffic control
5. Node Image: Use COS (Container-Optimized OS) for security updates

Cost Optimization

1. Spot Instances: Use for fault-tolerant workloads
2. Autoscaling: Configure appropriate min/max node counts
3. Right-sizing: Choose appropriate machine types
4. Preemptible Nodes: Alternative to spot instances for cost savings

Monitoring and Logging

The cluster is configured with:

• Google Cloud Logging integration
• Google Cloud Monitoring integration
• Cluster and node metrics collection

Troubleshooting

Common issues and solutions:

1. API not enabled: Enable required GCP APIs
2. Quota exceeded: Check GCP quotas for compute resources
3. Network connectivity: Verify firewall rules and network configuration
4. Authentication: Ensure proper GCP credentials are configured