Implementation:Kserve Kserve LocalModelNodeGroup CRD
| Knowledge Sources | |
|---|---|
| Domains | Kubernetes, CRD, Local Model Storage |
| Last Updated | 2026-02-13 00:00 GMT |
Overview
Concrete CRD definition for the LocalModelNodeGroup custom resource in the KServe serving API.
Description
This file contains the full CustomResourceDefinition for the LocalModelNodeGroup kind, produced by controller-gen v0.19.0. It belongs to the serving.kserve.io API group at version v1alpha1 and is a cluster-scoped resource. The CRD defines node groups with shared PersistentVolumeClaim and PersistentVolume specifications for local model storage, allowing cluster administrators to configure storage provisioning policies for nodes that host pre-cached ML models.
Usage
Apply this CRD during KServe installation to register the LocalModelNodeGroup API with the Kubernetes API server. Cluster administrators then create LocalModelNodeGroup resources to define groups of nodes with specific storage configurations for caching models locally, improving inference startup time by avoiding remote model downloads.
Code Reference
Source Location
- Repository: Kserve_Kserve
- File: config/crd/full/localmodel/serving.kserve.io_localmodelnodegroups.yaml
- Lines: 1-692
Signature
apiVersion: apiextensions.k8s.io/v1
kind: CustomResourceDefinition
metadata:
annotations:
controller-gen.kubebuilder.io/version: v0.19.0
name: localmodelnodegroups.serving.kserve.io
spec:
group: serving.kserve.io
names:
kind: LocalModelNodeGroup
listKind: LocalModelNodeGroupList
plural: localmodelnodegroups
singular: localmodelnodegroup
scope: Cluster
versions:
- name: v1alpha1
Import
kubectl apply -f config/crd/full/localmodel/serving.kserve.io_localmodelnodegroups.yaml
I/O Contract
Inputs
| Name | Type | Required | Description |
|---|---|---|---|
| apiVersion | string | Yes | Must be serving.kserve.io/v1alpha1
|
| kind | string | Yes | Must be LocalModelNodeGroup
|
| metadata | ObjectMeta | Yes | Standard Kubernetes object metadata (no namespace since cluster-scoped) |
| spec | LocalModelNodeGroupSpec | Yes | Storage configuration for the node group |
Key spec fields:
| Field | Type | Required | Description |
|---|---|---|---|
| spec.persistentVolumeClaimSpec | PersistentVolumeClaimSpec | Yes | PVC specification defining access modes, storage class, volume mode, data sources, resource requests/limits, and selectors for local model storage |
| spec.persistentVolumeSpec | PersistentVolumeSpec | Yes | PV specification defining the backing storage volume with support for various volume types (AWS EBS, Azure Disk, GCE PD, NFS, CSI, etc.) |
| spec.storageLimit | Quantity | Yes | Maximum storage capacity allocated for local model caching on nodes in this group |
Outputs
| Name | Type | Description |
|---|---|---|
| status.available | Quantity | Amount of storage currently available in the node group |
| status.used | Quantity | Amount of storage currently consumed by cached models in the node group |
Usage Examples
Create a LocalModelNodeGroup
apiVersion: serving.kserve.io/v1alpha1
kind: LocalModelNodeGroup
metadata:
name: gpu-nodes-storage
spec:
persistentVolumeClaimSpec:
accessModes:
- ReadWriteOnce
storageClassName: local-ssd
resources:
requests:
storage: 500Gi
persistentVolumeSpec:
accessModes:
- ReadWriteOnce
capacity:
storage: 500Gi
hostPath:
path: /mnt/model-cache
storageLimit: 500Gi
Related Pages
- New principle needed:
Kserve_Kserve_LocalModelNodeGroup_Specification