GKE Node Pool

Node Pool

Reference

• GKE / Deploy AI/ML workloads / Run GPUs in GKE Standard node pools

resource "google_container_node_pool" "<name>" {
  cluster = google_container_cluster.<name>.id
}

• location = <region|zone>
  • 클러스터가 위치한 리전 또는 존을 설정합니다.
• node_locations = ["<zone>"]
• node_config {}
  • machine_type = "<type>"
  • image_type = "COS_CONTAINERD|UBUNTU_CONTAINERD"
  • labels = {}
  • taint {}
    • key = "<key>"
    • value = "<value>"
    • effect = "NO_SCHEDULE|PREFER_NO_SCHEDULE|NO_EXECUTE"
  • guest_accelerator {}
    • Node의 GPU/TPU 정보를 설정합니다.
    • type = "<type>"
      • Node에 cloud.google.com/gke-accelerator: <type> 라벨이 자동으로 추가됩니다.
    • count = <count>
    • gpu_driver_installation_config {}
      • gpu_driver_version = "GPU_DRIVER_VERSION_UNSPECIFIED|INSTALLATION_DISABLED|DEFAULT|LATEST"
• network_config {}
  • additional_pod_network_configs
    • network = google_compute_network.<name>.name
    • subnetwork = google_compute_subnetwork.<name>.name