dev-l4-spot

Development Cluster with L4 Spot GPUs

Creates a minimal GKE cluster with a single L4 GPU spot node pool — ideal for development and testing SIE (Search Inference Engine) workloads at low cost.

What this example creates

Resource	Configuration
GKE cluster	Private nodes, Cloud NAT, Workload Identity
GPU node pool	1x NVIDIA L4 per node (g2-standard-8), spot VMs, scale 0-5
CPU node pool	e2-standard-4, scale 1-3 (system workloads)
Artifact Registry	Docker repository for SIE images
NAP	Node Auto-Provisioning enabled (auto-creates pools as needed)

Estimated cost: ~$0.50/hr when a GPU node is running. $0/hr when scaled to zero (only the GKE management fee applies).

Usage

export TF_VAR_project_id="your-gcp-project-id"

terraform init
terraform plan
terraform apply

After apply, deploy SIE via Helm:

# Configure kubectl
$(terraform output -raw kubectl_command)

# Install SIE (gateway, workers, KEDA, Prometheus, Grafana)
helm upgrade --install sie-cluster oci://ghcr.io/superlinked/charts/sie-cluster --version 0.3.4 \
  -f values-gke.yaml \
  --create-namespace -n sie \
  --set serviceAccount.annotations."iam\.gke\.io/gcp-service-account"="$(terraform output -raw workload_identity_annotation)"

Variables

Variable	Default	Description
project_id	— (required)	Your GCP project ID
region	us-central1	GCP region
cluster_name	sie-dev	Cluster name
create_artifact_registry	true	Create a Docker registry for SIE images
deployer_service_account	""	Service account email (for CI/CD; optional for interactive use)

Outputs

Output	Description
cluster_name	GKE cluster name
kubectl_command	Run this to configure kubectl
artifact_registry_url	URL for pushing Docker images
workload_identity_annotation	Annotation for Helm service account

Customizing

Change region:

export TF_VAR_region="europe-west4"

Use on-demand instead of spot (more reliable, higher cost):

Override gpu_node_pools in a terraform.tfvars file:

gpu_node_pools = [
  {
    name           = "l4-ondemand"
    machine_type   = "g2-standard-8"
    gpu_type       = "nvidia-l4"
    gpu_count      = 1
    min_node_count = 0
    max_node_count = 5
    spot           = false
  }
]

Prerequisites

1. GCP project with billing enabled
2. GPU quota for nvidia-l4 in your region (check: gcloud compute regions describe REGION --format="table(quotas.filter(metric:NVIDIA))")
3. APIs enabled: container.googleapis.com, compute.googleapis.com

Cleanup

terraform destroy