AWS EKS with Terraform

provider "aws" {
  region  = "eu-west-1"
}

module "vpc" {
  source  = "terraform-aws-modules/vpc/aws"
  version = "~> 5.1.2"

  name = "sombra-vpc"
  cidr = "10.0.0.0/16"
  azs  = ["eu-west-1a", "eu-west-1b"]

  private_subnets = ["10.0.101.0/24", "10.0.102.0/24"]
  public_subnets  = ["10.0.201.0/24", "10.0.202.0/24"]

  enable_nat_gateway   = true
  enable_dns_hostnames = true
  enable_dns_support   = true
}

module "eks_cluster" {
  source  = "terraform-aws-modules/eks/aws"
  version = "~> 20.31.6"

  cluster_name    = "sombra-eks-cluster"
  cluster_version = "1.32"

  vpc_id     = module.vpc.vpc_id
  subnet_ids = module.vpc.private_subnets

  cluster_endpoint_public_access = true

  enable_cluster_creator_admin_permissions = true

  cluster_compute_config = {
    enabled    = true
    node_pools = ["general-purpose"]
  }
}

###################################################################################
# Everything below here depends on the above kubernetes cluster already existing, #
#   so you may want to comment out the below resources during the first apply     #
###################################################################################

provider "kubernetes" {
  host                   = module.eks_cluster.cluster_endpoint
  cluster_ca_certificate = base64decode(module.eks_cluster.cluster_certificate_authority_data)
  exec {
    api_version = "client.authentication.k8s.io/v1beta1"
    args        = ["eks", "get-token", "--cluster-name", module.eks_cluster.cluster_name]
    command     = "aws"
  }
}

provider "helm" {
  kubernetes {
    host                   = module.eks_cluster.cluster_endpoint
    cluster_ca_certificate = base64decode(module.eks_cluster.cluster_certificate_authority_data)
    exec {
      api_version = "client.authentication.k8s.io/v1beta1"
      args        = ["eks", "get-token", "--cluster-name", module.eks_cluster.cluster_name]
      command     = "aws"
    }
  }
}

resource "kubernetes_manifest" "allow_gpu_node_pool" {
  manifest = {
    apiVersion = "karpenter.sh/v1"
    kind       = "NodePool"
    metadata = {
      name = "default"
    }
    spec = {
      template = {
        spec = {
          nodeClassRef = {
            group = "eks.amazonaws.com"
            kind  = "NodeClass"
            name  = "default"
          }
          requirements = [
            {
              key = "eks.amazonaws.com/instance-gpu-manufacturer"
              operator = "In"
              values = ["nvidia"]
            },
            # The A10G GPU has been tested to work well with our LLM Classifier
            {
              key = "eks.amazonaws.com/instance-gpu-name"
              operator = "In"
              values = ["a10g"]
            }
          ]
        }
      }
      limits = {
        cpu    = "16000"
        memory = "64Gi"
      }
    }
  }
}

resource "helm_release" "sombra" {
  name             = "sombra-servers"
  namespace        = "sombra"
  chart            = "sombra"
  create_namespace = true
  repository       = "https://transcend-io.github.io/helm-charts/"
  version          = "0.5.0"
  timeout          = 900 # 15 minutes for any helm operation
  values = [yamlencode({
    imageCredentials = {
      registry = "docker.transcend.io"
      username = "Transcend"
      password = "<TRANSCEND_API_KEY>"
    }
    replicaCount = 1
    envs = [
      {
        name  = "ORGANIZATION_URI"
        value = "<ORGANIZATION_URI>"
      },
      {
        name  = "SOMBRA_ID"
        value = "<SOMBRA_ID>"
      },
      {
        name  = "TRANSCEND_URL"
        value = "<TRANSCEND_URL>"
      },
      {
        name  = "LLM_CLASSIFIER_URL"
        value = "http://sombra-servers-llm-classifier.transcend.svc.cluster.local:6081/"
      }
    ]
    envs_as_secret = [
      {
        name  = "SOMBRA_REVERSE_TUNNEL_API_KEY"
        value = "<SOMBRA_REVERSE_TUNNEL_API_KEY>"
      },
      {
        name  = "JWT_ECDSA_KEY"
        value = "<JWT_ECDSA_KEY>"
      },
      {
        name  = "INTERNAL_KEY_HASH"
        value = "<INTERNAL_KEY_HASH>"
      }
    ]
    llm-classifier = { enabled = true }
    affinity = {
      nodeAffinity = {
        requiredDuringSchedulingIgnoredDuringExecution = {
          nodeSelectorTerms = [{
            matchExpressions = [{
              key      = "eks.amazonaws.com/instance-gpu-manufacturer"
              operator = "DoesNotExist"
            }]
          }]
        }
      }
    }
  })]
}