AgentRules.org
RulesStacksClientsAuditorPrunerSelf-HealRFCsBenchmarkCLICI/CDAuthorsSubmit
This specification is maintained by the AgentRules Working Group. Have enhancements or want to claim maintainership?Submit Specification →
devopscritical PriorityID: k8s-132-dra-structured-gpu-standard

Kubernetes 1.32 DRA Structured GPU Allocation Standard

Enforces Kubernetes 1.32 Dynamic Resource Allocation (DRA) Structured Parameters for deterministic multi-tenant GPU provisioning. Prevents legacy device plugin resource fragmentation and out-of-band driver re-initialization by replacing opaque string-keyed limits with resource.k8s.io/v1beta1 ResourceClaim specs and in-tree CEL selector constraints.

Prohibited Anti-Pattern: Opaque Device-Plugin Allocations and Runtime MPS Driver Re-Initialization Hacks

Export Ready-to-Drop Configuration

Select format below to copy code or download file directly.

---
description: Kubernetes 1.32 DRA Structured GPU Allocation Standard - Enforces Kubernetes 1.32 Dynamic Resource Allocation (DRA) Structured Parameters for deterministic multi-tenant GPU provisioning. Prevents legacy device plugin resource fragmentation and out-of-band driver re-initialization by replacing opaque string-keyed limits with resource.k8s.io/v1beta1 ResourceClaim specs and in-tree CEL selector constraints.
globs: deploy/**/*.{yaml,yml}, helm/**/*.{yaml,yml}, src/**/*.{ts,tsx}
alwaysApply: true
verified_by: "agentrules.org"
license: "CC-BY-4.0"
source: "https://agentrules.org/rules/k8s-132-dra-structured-gpu-standard"
---

# Kubernetes 1.32 DRA Structured GPU Allocation Standard

> **Severity**: CRITICAL | **Category**: devops
> **Anti-Pattern Warning**: Opaque Device-Plugin Allocations and Runtime MPS Driver Re-Initialization Hacks

## System Constraints
You are a Principal DevOps and Systems Architect enforcing the Kubernetes 1.32 Dynamic Resource Allocation (DRA) Structured GPU Standard. Mandate resource.k8s.io/v1beta1 ResourceClaims, DeviceClasses, and Structured Parameters via CEL expressions. Reject any legacy device plugin limits ('nvidia.com/gpu'), opaque environment-based MPS slicing, or driver re-initialization hacks that bypass the native Kubernetes scheduler.

## Strict Guidelines
- Engineers MUST ALWAYS declare GPU requirements via 'resource.k8s.io/v1beta1' ResourceClaim or ResourceClaimTemplate objects instead of 'spec.containers[*].resources.limits["nvidia.com/gpu"]'.
- Engineers MUST NOT mutate NVIDIA driver state or reinvoke runtime partitioning daemons (e.g., dynamic nvidia-smi reconfiguration or MPS server launches) inside container entrypoints.
- Workloads requiring fractional GPU compute or isolated VRAM MUST define CEL-based structured selectors matching 'ResourceSlice' attributes published by the DRA driver.
- Pods requiring structured GPU resources MUST bind claims using 'spec.resourceClaims[*]' and reference them inside 'spec.containers[*].resources.claims[*]'.
- DeviceClasses MUST define rigorous structured configuration requests so that kube-scheduler can evaluate hardware feasibility in-tree without out-of-band driver calls.

## Bad vs Good Implementations
### CDK8s GPU Provisioning: Legacy Device Plugin vs. K8s 1.32 Structured DRA

#### Bad Antipattern (AI Common Hallucination)
> **Why it fails**: Legacy device plugin limits ('nvidia.com/gpu') treat accelerators as monolithic units or rely on fragile, out-of-band runtime driver hacks (like manual CUDA_MPS_PINNED_DEVICE_MEM_LIMIT) that can fail at runtime, cause driver memory leaks, and bypass kube-scheduler's bin-packing state.
```typescript
import { Construct } from 'constructs';
import { Chart, ChartProps } from 'cdk8s';
import { KubePod } from 'cdk8s-plus-30/lib/imports/k8s';

export class LegacyGpuWorkload extends Chart {
  constructor(scope: Construct, id: string, props?: ChartProps) {
    super(scope, id, props);

    // ANTI-PATTERN: Using deprecated opaque integer device limits with out-of-band MPS env vars
    new KubePod(this, 'GpuInferencePod', {
      metadata: { name: 'llm-infer-legacy' },
      spec: {
        containers: [{
          name: 'inference',
          image: 'vllm/vllm-openai:v0.6.0',
          env: [
            // Flawed attempt to slice GPU via runtime driver hacking
            { name: 'CUDA_MPS_PINNED_DEVICE_MEM_LIMIT', value: '16G' },
            { name: 'CUDA_VISIBLE_DEVICES', value: '0' }
          ],
          resources: {
            limits: {
              'nvidia.com/gpu': '1' // Opaque, non-deterministic scheduling lock
            }
          }
        }]
      }
    });
  }
}
```

#### Production Standard (Correct)
> **Standard**: Leverages Kubernetes 1.32 DRA Structured Parameters (`resource.k8s.io/v1beta1`). The scheduler natively matches hardware specifications (memory and model) using in-tree CEL expressions against `ResourceSlice` data, guaranteeing deterministic fractional allocation without container-level driver reloads.
```typescript
import { Construct } from 'constructs';
import { Chart, ChartProps } from 'cdk8s';
import { ApiObject } from 'cdk8s';

export class StructuredDraGpuWorkload extends Chart {
  constructor(scope: Construct, id: string, props?: ChartProps) {
    super(scope, id, props);

    // Invariant: Use K8s 1.32 DRA Structured Parameters API
    const claimTemplate = new ApiObject(this, 'GpuClaimTemplate', {
      apiVersion: 'resource.k8s.io/v1beta1',
      kind: 'ResourceClaimTemplate',
      metadata: { name: 'h100-fractional-claim' },
      spec: {
        spec: {
          devices: {
            requests: [{
              name: 'gpu-slice',
              deviceClassName: 'gpu.nvidia.com',
              selectors: [{
                cel: {
                  // Evaluated directly in-tree by kube-scheduler
                  expression: "device.attributes['gpu.nvidia.com'].memory >= quantity('16Gi') && device.attributes['gpu.nvidia.com'].model == 'H100'"
                }
              }]
            }]
          }
        }
      }
    });

    new ApiObject(this, 'GpuInferencePod', {
      apiVersion: 'v1',
      kind: 'Pod',
      metadata: { name: 'llm-infer-dra' },
      spec: {
        resourceClaims: [{
          name: 'pod-gpu',
          resourceClaimTemplateName: claimTemplate.name
        }],
        containers: [{
          name: 'inference',
          image: 'vllm/vllm-openai:v0.6.0',
          resources: {
            claims: [{
              name: 'pod-gpu'
            }]
          }
        }]
      }
    });
  }
}
```
Recommended MCP Servers (mcp-serverlist.com)

Equip agent with system tools to verify code for this stack.

Kubernetes Cluster Inspector MCP
Inspects active Kubernetes API schemas and verifies CRD support for resource.k8s.io/v1beta1 Dynamic Resource Allocation.
Recommended Agent Skills (askills.dev)

Ready-to-use executable skill functions tailored to this stack.

DRA ResourceSlice Validator
Validates cluster ResourceSlices and CEL expressions against published device parameters in Kubernetes 1.32+.
Architectural Specification & Invariant Rationale

Architectural Invariant Design: Kubernetes 1.32 DRA Structured GPU Allocation Standard

Kubernetes 1.32 DRA Structured GPU Allocation Standard defines an industrial-grade behavioral standard for devops engineering. Enforces Kubernetes 1.32 Dynamic Resource Allocation (DRA) Structured Parameters for deterministic multi-tenant GPU provisioning. Prevents legacy device plugin resource fragmentation and out-of-band driver re-initialization by replacing opaque string-keyed limits with resource.k8s.io/v1beta1 ResourceClaim specs and in-tree CEL selector constraints. Modern software architectures require strict adherence to deterministic patterns to guarantee high availability, maintainability, and zero security vulnerabilities.

Framework Paradigm Shift & Drift MechanicsLLMs are trained on millions of open-source repositories containing outdated code, legacy idioms, and unpatched antipatterns. When coding assistants generate implementations for this stack, they default to common statistical tokens: Opaque Device-Plugin Allocations and Runtime MPS Driver Re-Initialization Hacks Without an explicit system instruction file, developers spend hours debugging silent failures.
Failure Modes & Production Blast Radius

Hazard Analysis: What Breaks Without This Rule

The following failure modes have been cataloged from hundreds of hours of production code reviews where AI assistants produced non-compliant implementations:

Failure ModeAI Symptom (The Anti-Pattern)SeverityProduction Impact & Remedy
AI Hallucination & Deprecation DriftOpaque Device-Plugin Allocations and Runtime MPS Driver Re-Initialization HacksCritical
Violating this industrial standard introduces regression bugs, security risks, and unexpected runtime failures in production.
Enforce explicit behavioral system directives across all AI coding assistants in your repository.
Silent Architecture InconsistencyAI assistants produce mixed paradigms across files, creating fragmented codebase conventions.Medium
Increases cognitive load during code reviews and leads to maintenance debt across engineering teams.
Pin project conventions in .cursor/rules, CLAUDE.md, and AGENTS.md so all team agents follow identical guardrails.
Architectural Directives in Detail

Step-by-Step Production Directives

Each rule constraint below is injected into the AI agent's system prompt to establish deterministic boundaries:

1

Engineers MUST ALWAYS declare GPU requirements via 'resource.k8s.io/v1beta1' ResourceClaim or ResourceClaimTemplate objects instead of 'spec.containers[*].resources.limits["nvidia.com/gpu"]'.

Engineers MUST ALWAYS declare GPU requirements via 'resource.k8s.io/v1beta1' ResourceClaim or ResourceClaimTemplate objects instead of 'spec.containers[*].resources.limits["nvidia.com/gpu"]'.

Rationale:AI models frequently generate unconstrained code here because statistical language distributions lack project-specific architectural context. Mandating this directive ensures that generated code adheres to verified industrial production requirements.
Enforcement:Integrated into the Agent Pack rule file. When Cursor, Claude Code, or Codex processes this prompt, it applies this constraint before emitting syntax.
2

Engineers MUST NOT mutate NVIDIA driver state or reinvoke runtime partitioning daemons (e.g., dynamic nvidia-smi reconfiguration or MPS server launches) inside container entrypoints.

Engineers MUST NOT mutate NVIDIA driver state or reinvoke runtime partitioning daemons (e.g., dynamic nvidia-smi reconfiguration or MPS server launches) inside container entrypoints.

Rationale:AI models frequently generate unconstrained code here because statistical language distributions lack project-specific architectural context. Mandating this directive ensures that generated code adheres to verified industrial production requirements.
Enforcement:Integrated into the Agent Pack rule file. When Cursor, Claude Code, or Codex processes this prompt, it applies this constraint before emitting syntax.
3

Workloads requiring fractional GPU compute or isolated VRAM MUST define CEL-based structured selectors matching 'ResourceSlice' attributes published by the DRA driver.

Workloads requiring fractional GPU compute or isolated VRAM MUST define CEL-based structured selectors matching 'ResourceSlice' attributes published by the DRA driver.

Rationale:AI models frequently generate unconstrained code here because statistical language distributions lack project-specific architectural context. Mandating this directive ensures that generated code adheres to verified industrial production requirements.
Enforcement:Integrated into the Agent Pack rule file. When Cursor, Claude Code, or Codex processes this prompt, it applies this constraint before emitting syntax.
4

Pods requiring structured GPU resources MUST bind claims using 'spec.resourceClaims[*]' and reference them inside 'spec.containers[*].resources.claims[*]'.

Pods requiring structured GPU resources MUST bind claims using 'spec.resourceClaims[*]' and reference them inside 'spec.containers[*].resources.claims[*]'.

Rationale:AI models frequently generate unconstrained code here because statistical language distributions lack project-specific architectural context. Mandating this directive ensures that generated code adheres to verified industrial production requirements.
Enforcement:Integrated into the Agent Pack rule file. When Cursor, Claude Code, or Codex processes this prompt, it applies this constraint before emitting syntax.
5

DeviceClasses MUST define rigorous structured configuration requests so that kube-scheduler can evaluate hardware feasibility in-tree without out-of-band driver calls.

DeviceClasses MUST define rigorous structured configuration requests so that kube-scheduler can evaluate hardware feasibility in-tree without out-of-band driver calls.

Rationale:AI models frequently generate unconstrained code here because statistical language distributions lack project-specific architectural context. Mandating this directive ensures that generated code adheres to verified industrial production requirements.
Enforcement:Integrated into the Agent Pack rule file. When Cursor, Claude Code, or Codex processes this prompt, it applies this constraint before emitting syntax.
Technical Questions & Answers

Frequently Asked Technical Questions

Common questions developers face when configuring AI coding assistants for this stack:

Q:Why is Kubernetes 1.32 DRA Structured GPU Allocation Standard critical for AI coding assistants?

Enforces Kubernetes 1.32 Dynamic Resource Allocation (DRA) Structured Parameters for deterministic multi-tenant GPU provisioning. Prevents legacy device plugin resource fragmentation and out-of-band driver re-initialization by replacing opaque string-keyed limits with resource.k8s.io/v1beta1 ResourceClaim specs and in-tree CEL selector constraints. Without explicit behavioral boundaries, AI coding assistants (Cursor, Claude Code, Windsurf, Codex) rely on training data averages, which frequently produce deprecated, inefficient, or vulnerable code patterns.

Q:How does adopting this rule prevent AI hallucinations?

This rule explicitly declares what the AI must avoid (Opaque Device-Plugin Allocations and Runtime MPS Driver Re-Initialization Hacks) and prescribes verified positive examples. Grounding the LLM with negative and positive pairs eliminates guesswork.

Q:Which AI coding environments are supported?

This specification is multi-platform compatible: Cursor (.cursor/rules/*.mdc & .cursorrules), Anthropic Claude Code (CLAUDE.md), Windsurf (.windsurfrules), Codex / Cline (AGENTS.md), and GitHub Copilot (.github/copilot-instructions.md).

Q:Can I combine this rule with other technology stacks?

Yes. Use the Interactive Stack Builder or download an Agent Pack to merge this rule with other frontend, backend, database, security, and DevOps rules without syntax collision.

Open Specification · CC BY 4.0·Free for Personal, Team & Commercial In-Repo Use

Citation & Open Source Attribution Guidelines

This specification is licensed under Creative Commons Attribution 4.0 International (CC BY 4.0). You are completely free to adapt and share this rule in your workspace. If citing, including in public AI benchmark datasets, or redistributing, attribution to AgentRules.org is required.

Related DEVOPS Rules

View all rules →
devops
86

Kubernetes Gateway API v1 HTTPRoute Canary and ReferenceGrant Invariants

Enforces strict conformance to Kubernetes Gateway API v1 specifications during migration from legacy Ingress controllers. Mandates explicit ReferenceGrant resources for cross-namespace routing, strict relative integer weighting for HTTPRoute canary releases, and explicit backendRef port declarations.

devops
118

ArgoCD ApplicationSet Matrix Generator Invariants & Cascade Deletion Guards

Enforces strict structural constraints on ArgoCD ApplicationSet Matrix generators combining cluster selectors and Git directories. Prevents catastrophic multi-cluster cascade pruning, accidental secret exposure through parameter interpolation, and runaway Application CR explosion via explicit sync policies, preservation finalizers, and parameter validation.

devops
40

Karpenter Resilient Topology, Interruption Queue, and Consolidation Budget Standard

Enforces deterministic disruption budgets, multi-instance-type diversification, and event-driven SQS interruption queue handling across Karpenter NodePools and EC2NodeClasses. Eliminates cascading eviction loops, instance starvation, and availability downtime during AWS Spot rebalances and underutilized node consolidation.

1 Rule in Stack

Kubernetes