GridGain 9 on AWS EKS - Terraform Deployment

This Terraform configuration deploys a GridGain 9 cluster on AWS EKS with dedicated node groups.

Architecture

• Default Node Group: 2x m7g.medium instances for system workloads (CoreDNS, EBS CSI driver, etc.)
• GridGain Node Group: 3x m7gd.2xlarge instances dedicated to GridGain pods (tainted)

Prerequisites

1. Terraform >= 1.5.0
2. AWS CLI configured with appropriate credentials
3. kubectl for cluster management
4. Helm (optional, for manual operations)
5. GridGain License file (JSON format)

Configuration

1. Create terraform.tfvars

cp terraform.tfvars.example terraform.tfvars
# Edit terraform.tfvars with your values

2. Variables

Variable	Default	Description
aws_region	us-east-1	AWS region for deployment
aws_profile	null	AWS CLI profile (uses default if not set)
cluster_name	gg9-eks	EKS cluster name
cluster_version	1.30	Kubernetes version
node_instance_type	m7gd.2xlarge	Instance type for GridGain nodes
node_desired_size	3	Number of GridGain nodes
node_min_size	3	Minimum GridGain nodes
node_max_size	6	Maximum GridGain nodes
gg9_namespace	gridgain	Kubernetes namespace
gg9_chart_version	1.1.1	GridGain Helm chart version
gg9_license_secret_arn	(required)	ARN of AWS Secrets Manager secret containing the GridGain license

3. License Setup (AWS Secrets Manager)

Store your GridGain license in AWS Secrets Manager:

# Create the secret (replace with your actual license content)
aws secretsmanager create-secret \
  --name gridgain-license \
  --secret-string file://gridgain-license.json \
  --region us-east-1

# Get the ARN for terraform.tfvars
aws secretsmanager describe-secret --secret-id gridgain-license --query 'ARN' --output text

Then add the ARN to your terraform.tfvars:

gg9_license_secret_arn = "arn:aws:secretsmanager:us-east-1:123456789012:secret:gridgain-license-AbCdEf"

4. GridGain Configuration (gg9-values.yaml)

Key sections to customize:

Node Resources

resources:
  requests:
    cpu: "4"
    memory: "24Gi"
  limits:
    cpu: "7"
    memory: "28Gi"

Storage (Local NVMe)

# Local NVMe for all data (maximum performance)
extraVolumes:
  - name: nvme-data
    hostPath:
      path: /mnt/nvme

extraVolumeMounts:
  - name: nvme-data
    mountPath: /data

gridgainWorkDir: /data

Durability is provided by RAFT replication across 3 nodes.

Node Finder (must match your Helm release name)

nodeFinder {
  type = STATIC
  netClusterNodes = [
    "gg9-gridgain9-headless:3344"  # Format: <release-name>-gridgain9-headless:3344
  ]
}

Deployment

Initial Setup

# Configure AWS profile
aws configure --profile your-profile-name

# Initialize Terraform
terraform init

# Review changes
terraform plan

# Deploy
terraform apply

Verify Deployment

# Check nodes
kubectl get nodes

# Check GridGain pods
kubectl get pods -n gridgain

# Check cluster status
kubectl exec -it gg9-gridgain9-0 -n gridgain -- \
  /opt/gridgain9cli/bin/gridgain9 cluster status

# Check cluster topology
kubectl exec -it gg9-gridgain9-0 -n gridgain -- \
  /opt/gridgain9cli/bin/gridgain9 cluster topology physical --url=http://localhost:10300

Connecting to GridGain

Option 1: Load Balancer

The deployment provisions an AWS Classic Load Balancer for external client access:

# Get the Load Balancer hostname
kubectl get svc gg9-gridgain9-client -n gridgain -o jsonpath='{.status.loadBalancer.ingress[0].hostname}'

JDBC Connection String:

jdbc:ignite:thin://<LB_HOSTNAME>:10800

Note: The Load Balancer may take 2-3 minutes to provision and become healthy after deployment.

The client service uses externalTrafficPolicy: Local to avoid extra network hops and preserve client source IPs.

Option 2: Port Forward (Development)

# Start port forwarding (run in background or separate terminal)
kubectl port-forward svc/gg9-gridgain9-headless 10800:10800 -n gridgain

# For REST API access
kubectl port-forward svc/gg9-gridgain9-headless 10300:10300 -n gridgain

JDBC Connection String:

jdbc:ignite:thin://localhost:10800

Example: Connect with DBeaver/DataGrip

Using Load Balancer:

1. Get LB hostname: kubectl get svc gg9-gridgain9-client -n gridgain
2. Add new connection in your SQL client
3. Use driver: Apache Ignite (or GridGain)
4. Host: <LB_HOSTNAME>
5. Port: 10800

Using Port Forward:

1. Start port-forward: kubectl port-forward svc/gg9-gridgain9-headless 10800:10800 -n gridgain
2. Host: localhost
3. Port: 10800

Cleanup

# Destroy all resources
terraform destroy

Troubleshooting

Pods Pending - Insufficient Memory

Reduce memory requests in gg9-values.yaml or use larger instance types.

Init Job Failing - License Not Found

Ensure license file path is correct and file contains valid JSON license.

Nodes Not Discovering Each Other

Check nodeFinder.netClusterNodes matches your Helm release name:

• Release name gg9 → gg9-gridgain9-headless:3344
• Release name my-release → my-release-gridgain9-headless:3344

CoreDNS/EBS CSI Pending

Ensure the default node group exists and has available capacity.

Multi-Region & Data Center Replication (DCR)

For multi-region deployments with cross-cluster replication, see the DCR Guide.

Highlights:

• VPC Peering vs Transit Gateway connectivity options
• Using the client LoadBalancer service for stable DCR endpoints
• Bidirectional replication configuration
• Security group and routing setup

File Structure

├── main.tf                    # EKS cluster, node groups, addons
├── vpc.tf                     # VPC configuration
├── variables.tf               # Input variables
├── outputs.tf                 # Output values
├── gg9-helm.tf                # GridGain Helm release + license secret
├── gg9-values.yaml            # GridGain Helm values
├── terraform.tfvars.example   # Example variables file
└── dcr/                       # Multi-region DCR setup (see dcr/README.md)

Instance Types with Local NVMe

Use m7gd.* or r7gd.* instances which include local NVMe:

Instance	vCPU	Memory	NVMe Storage
m7gd.xlarge	4	16 GiB	1x 118 GiB
m7gd.2xlarge	8	32 GiB	1x 237 GiB
m7gd.4xlarge	16	64 GiB	1x 474 GiB

Note: Local NVMe is ephemeral. Durability is provided by RAFT replication across 3 nodes.