#Add Custom Devices Using ConfigMap

#Introduction

• Implements standardized definition and management of Kubernetes custom resources through ConfigMap, addressing:
• Unified management of custom resource specifications to prevent configuration fragmentation
• Standardized resource definition format for better maintainability
• Multi-language description support and default value configuration
• Suitable for scenarios requiring Kubernetes resource model extension (e.g., GPU resource management), providing a standardized resource definition framework

#Features

• Single-key resource definition specification
• Multi-key associated resource definition
• Standardized resource request interface
• Chinese/English bilingual description support
• Resource default value configuration mechanism

#Advantages

• Extensibility: Resource group management through labels
• Security: Namespace isolation (kube-public)
• Stability: Enforced format validation rules
• Maintainability: Unified metadata label specifications

#Function Module 1: ConfigMap Authoring Specifications

#Core Rules

1. Single Responsibility Principle: One ConfigMap per key definition

2. Namespace: Fixed to namespace=kube-public

3. Naming Convention:

   cf-crl-{customName}-{keyName}
   • cf-crl: Fixed prefix
   • customName: Custom valid name
   • keyName: Key identifier (special characters replaced with '-')

4. Label Requirements:

   labels:
     features.alauda.io/type: CustomResourceLimitation  # Fixed value
     features.alauda.io/group: {resource-group}         # e.g., gpu-manager
     features.alauda.io/enabled: "true"               # Activation flag
   

#Parameter Specification

#Function Module 2: Resource Value Definition

#Single Key Example

apiVersion: v1
kind: ConfigMap
metadata:
  name: cf-crl-gpu-manager-vcuda-core
  namespace: kube-public
  labels:
    features.alauda.io/type: CustomResourceLimitation
    features.alauda.io/group: gpu-manager
    features.alauda.io/enabled: "true"
data:
  key: "tencent.com/vcuda-core"       # Resource key
  dataType: "integer"                # Value type
  defaultValue: "20"                 # Default value
  descriptionZh: "" # Chinese description
  descriptionEn: "GPU vcore count, 100 virtual cores equal 1 physical GPU core" # English description
  group: "gpu-manager"               # Resource group
  limits: "optional"                 # Limits field policy
  requests: "disabled"               # Requests field policy

#Multi-key Association

metadata:
  name: cf-crl-gpu-manager-vcuda-core
  labels: [same group labels]

metadata:
  name: cf-crl-gpu-manager-vcuda-memory
  labels: [same group labels]  # Association through identical labels

#Policy Specification