Implementation:Kserve Kserve LLMInferenceService Minimal CRD

Knowledge Sources	Kserve_Kserve
Domains	Kubernetes, LLM Inference, CRD
Last Updated	2026-02-13 00:00 GMT

Overview

This file defines the minimal CustomResourceDefinition (CRD) for the LLMInferenceService custom resource (shortName: llmisvc) in the serving.kserve.io API group.

Description

Auto-generated by controller-gen (v0.19.0), this CRD provides a smaller-footprint alternative to the full LLMInferenceService CRD. The v1alpha1 version uses x-kubernetes-preserve-unknown-fields for both spec and status, while the v1alpha2 version includes partial schema definitions for model, parallelism, router, and template fields. The CRD includes printer columns for URL, Ready, Reason, and Age, making kubectl get llmisvc output informative.

Usage

Use this minimal CRD for development or constrained environments where CRD size is a concern. Apply it to the cluster before creating any LLMInferenceService objects. The short name llmisvc allows convenient access via kubectl get llmisvc.

Code Reference

Source Location

• Repository: Kserve_Kserve
• File: config/crd/minimal/llmisvc/serving.kserve.io_llminferenceservices.yaml

Signature

apiVersion: apiextensions.k8s.io/v1
kind: CustomResourceDefinition
metadata:
  annotations:
    controller-gen.kubebuilder.io/version: v0.19.0
  name: llminferenceservices.serving.kserve.io
spec:
  group: serving.kserve.io
  names:
    kind: LLMInferenceService
    listKind: LLMInferenceServiceList
    plural: llminferenceservices
    shortNames:
    - llmisvc
    singular: llminferenceservice
  scope: Namespaced
  versions:
  - additionalPrinterColumns:
    - jsonPath: .status.url
      name: URL
      type: string
    - jsonPath: .status.conditions[?(@.type=='Ready')].status
      name: Ready
      type: string
    - jsonPath: .status.conditions[?(@.type=='Ready')].reason
      name: Reason
      type: string
    - jsonPath: .metadata.creationTimestamp
      name: Age
      type: date
    name: v1alpha1
    schema:
      openAPIV3Schema:
        properties:
          spec:
            type: object
            x-kubernetes-preserve-unknown-fields: true
          status:
            type: object
            x-kubernetes-preserve-unknown-fields: true
        type: object
    served: true
    storage: false
    subresources:
      status: {}
  - name: v1alpha2
    # ... includes partial schema for baseRefs, model, parallelism, router, template

Import

kubectl apply -f config/crd/minimal/llmisvc/serving.kserve.io_llminferenceservices.yaml

I/O Contract

Field	Type	Description
apiVersion	string	Must be apiextensions.k8s.io/v1
kind	string	Must be CustomResourceDefinition
spec.group	string	API group: serving.kserve.io
spec.names.kind	string	LLMInferenceService
spec.names.shortNames	array	["llmisvc"]
spec.scope	string	Namespaced

Printer Columns

Column	JSONPath	Type
URL	.status.url	string
Ready	.status.conditions[?(@.type=='Ready')].status	string
Reason	.status.conditions[?(@.type=='Ready')].reason	string
Age	.metadata.creationTimestamp	date
URLs (priority 1)	.status.addresses[*].url	string

Versions

Version	Storage	Served	Schema Detail
v1alpha1	false	true	Minimal schema with x-kubernetes-preserve-unknown-fields for spec and status
v1alpha2	true	true	Partial schema with baseRefs, model, parallelism, router, template, prefill, worker

v1alpha2 Key Spec Fields

Field	Type	Description
spec.model	object	Model specification with uri (required), name, and optional lora adapters
spec.parallelism	object	Parallelism configuration: data, dataLocal, tensor, expert
spec.replicas	integer	Number of replicas for the inference service
spec.router	object	Router configuration with scheduler, route, and gateway
spec.template	object	Pod template for the main serving containers
spec.prefill	object	Separate prefill pool configuration for disaggregated inference
spec.worker	object	Worker pod template for data-parallel worker nodes

Usage Examples

Register the CRD and create an LLMInferenceService:

# Apply the minimal CRD definition
kubectl apply -f config/crd/minimal/llmisvc/serving.kserve.io_llminferenceservices.yaml

# Verify the CRD is registered
kubectl get crd llminferenceservices.serving.kserve.io

# List LLMInferenceService resources using the short name
kubectl get llmisvc

Related Pages

• Kserve_Kserve_LLMInferenceServiceConfig_Minimal_CRD - Companion minimal CRD for the LLMInferenceServiceConfig resource
• Kserve_Kserve_Gateway_Inference_Extension_CRDs - Gateway Inference Extension CRDs used for routing
• Kserve_Kserve_DeepSeek_R1_PD_DeepEP_HT_Sample - Sample LLMInferenceService deployment for DeepSeek-R1
• Kserve_Kserve_DeepSeek_R1_PD_DeepEP_Pplx_Sample - Sample LLMInferenceService deployment with hybrid backends