Modelplane Modelplane docs
Version
main Latest

GKECluster Custom Resource

On this page

A GKECluster provisions a GKE cluster with dedicated node pools for GPU inference and system workloads. It outputs secrets containing the cluster kubeconfig and a GCP service account key that consumers can use to target the cluster.

#Metadata

API version
infrastructure.modelplane.ai/v1alpha1
Kind
GKECluster
Scope
Namespaced
Short names
gke

#Example

Manifest
apiVersion: infrastructure.modelplane.ai/v1alpha1
kind: GKECluster
metadata:
  name: inference-us-central
  namespace: platform
spec:
  project: my-gcp-project
  region: us-central1
  nodePools:
    - name: system
      role: System
      machineType: e2-standard-4
      nodeCount: 2
    - name: gpu-h100
      role: GPU
      machineType: a3-highgpu-8g
      diskSizeGb: 500
      nodeCount: 0
      maxNodeCount: 4
      zones: [us-central1-a]
      gpu:
        acceleratorType: nvidia-h100-80gb
        acceleratorCount: 8

#Spec

GKEClusterSpec defines the desired state of GKECluster.

# kubernetesVersion optional string 1–16 chars default: 1.35

GKE cluster Kubernetes version. Must be a version supported by the REGULAR release channel.

# networking optional object

VPC networking configuration. Defaults are suitable for standalone clusters. Override when VPC-peering multiple clusters to avoid CIDR collisions.

# nodeCidr optional string ≤ 18 chars default: 10.0.0.0/24

Primary IP range for nodes.

# podCidr optional string ≤ 18 chars default: 10.1.0.0/16

Secondary IP range for pods.

# serviceCidr optional string ≤ 18 chars default: 10.2.0.0/16

Secondary IP range for services.

# nodePools required object[] 1–8 items
# diskSizeGb optional integer 10–65536 default: 100

Boot disk size in GB.

# gpu optional object

GPU configuration. Required when role is GPU.

# acceleratorCount optional integer 1–16 default: 1

Number of GPUs to attach per node.

# acceleratorType required string 1–63 chars

GPU accelerator type (e.g. nvidia-tesla-a100, nvidia-h100-80gb, nvidia-l4). Attached to each node along with acceleratorCount.

# machineType required string 1–63 chars

GCE machine type (e.g. e2-standard-4, a2-highgpu-8g, g2-standard-48).

# maxNodeCount optional integer ≤ 1000 default: 8

Maximum number of nodes for autoscaling.

# minNodeCount optional integer ≤ 1000

Minimum number of nodes for autoscaling. Set to 1 or higher for pools that must always be available.

# name required string 1–40 chars

Unique name for this node pool. Used as a suffix in the GKE NodePool resource name.

# nodeCount optional integer ≤ 1000 default: 1

Initial number of nodes.

# role required enum: System | GPU

Determines what workloads this pool runs. System pools host controllers, gateways, and infrastructure. GPU pools host inference workloads and are tainted to exclude non-GPU pods.

# zones optional string[] 1–8 items 1–63 chars
# project required string 6–30 chars

GCP project ID where the cluster will be created.

# region required string 1–32 chars

GCP region for the cluster (e.g. us-central1, europe-west4).

#Status

# cache optional object

Observed ModelCache RWX storage state.

# storageClassName optional string ≤ 253 chars

Name of the Modelplane-managed ReadWriteMany StorageClass composed on this cluster for ModelCache PVCs. ModelCache reads this to target the cache PVC.

# secrets optional object[]
# key required string ≤ 253 chars

Key within the Secret that holds the credential data.

# name required string ≤ 253 chars

Name of the Secret.

# type required enum: Kubeconfig | GCPServiceAccountKey

The type of credential this secret contains. Kubeconfig contains a kubeconfig file with the cluster endpoint and CA certificate. GCPServiceAccountKey contains a GCP service account JSON key that can authenticate to the cluster via GKE IAM.