• VPC CIDR should be atleast /20 for the VPC
• Min 2 availability zone and /24 for private subnets

This is needed to ensure around 250 instances and 4096 pods can be run in the Kubernetes cluster. If we expect the scale to be higher, the subnet range should be increased.

A NAT Gateway must be connected in the VPC and the route tables should allow outbound internet access for private subnets through this NAT gateway.

1. public.ecr.aws
2. quay.io
3. ghcr.io
4. docker.io/truefoundrycloud
5. docker.io/natsio
6. nvcr.io
7. registry.k8s.io

This is to download docker images for TrueFoundry, ArgoCD, NATS, GPU operator, ArgoRollouts, ArgoWorkflows, Istio, Keda.

Set of endpoints (preferably wildcard) to point to the deployments being made. Something like *.internal.example.com, *.external.example.com. An ACM certificate with the chosen domains as SAN is required in the same region

When developers deploy their services, they will need to access the endpoints of their services to test it out or call from other services. This is why we need the DNS along with TLS on the compute plane. Its better if we can make it a wildcard since then developers can deploy services like service1.internal.example.com, service2.internal.example.com

We need to have a certificate for the domains listed above. The certificate ARN will be passed to the Istio Ingress config.

If you have a certificate from some other source, that can also work by creating a secret with the certificate in istio-system namespace.

GPU

If you are planning to use GPU machines, make sure you have quotas for:

• G and VT Spot/On-demand Instances
• P Spot/On-demand Instance Requests

Inferentia (Optional)

• If you are planning to use Inferentia machines, make sure you have quota for Inferentia Spot/On-demand machines.

This is to make sure that TrueFoundry can bring up the instances as requested by developers. A request needs to be raised to AWS for increasing the limits for instances in case we don't have quotas. You can check and increase your quotas at AWS EC2 service quotas

1. sts must be enabled for the user which is being used to create the cluster.
2. User must have the list of permissions listed below

See Enabling STS in a region

• Minimum 2 private subnets in different availability zone with min CIDR /24
• Tags should be present on the VPC as described below
• NAT gateway for private subnets
• Minimum 1 public subnet for a public load balancer if endpoints are to be exposed to internet. Auto-assign IP address must be enabled. Min CIDR /28
• DNS support and DNS hostnames must be enabled for your VPC

This is needed to ensure around 250 instances and 4096 pods can be run in the Kubernetes cluster. If we expect the scale to be higher, the subnet range should be increased.

A NAT Gateway must be connected in the VPC and the route tables should allow outbound internet access for private subnets through this NAT gateway.

"kubernetes.io/cluster/${clusterName}": "shared"
"subnet": "private"
"kubernetes.io/role/internal-elb": "1"

"kubernetes.io/cluster/${clusterName}": "shared"
"subnet": "public"
"kubernetes.io/role/elb": "1"

"karpenter.sh/discovery": "${clusterName}"

This tag is required for Karpenter to discover and manage node provisioning for the cluster.

CPU

• All Standard (A, C, D, H, I, M, R, T, Z) Spot/On-demand must have min 4vCPU and 8 GB RAM
• At least 2 nodes should be available for system components

• EKS version 1.30 or higher

Required for compatibility with TrueFoundry components and latest security features. Newer versions provide better performance and stability.

• EBS CSI Driver must be installed
  • Installation Guide for EBS CSI Driver
• EFS CSI Driver (if using shared storage)
  • Installation Guide for EFS CSI Driver

Required for persistent volume provisioning and shared storage support.

• AWS Load Balancer Controller version 2.12.0 or higher
  • Installation Guide for AWS Load Balancer Controller
• Appropriate IAM roles for service account (IRSA)

Required for Ingress and Service type LoadBalancer support.