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-dra-structured-parameter-standard

Kubernetes Dynamic Resource Allocation (DRA) Structured Parameter Invariant

Enforces deterministic accelerator provisioning using Kubernetes Dynamic Resource Allocation (DRA) structured parameters and Common Device Interface (CDI) descriptors instead of legacy opaque integer limits. Prevents cross-workload PCIe/NVLink contention, memory fragmentation, and noisy-neighbor stalls in heterogeneous AI/ML clusters by mandating CEL-based hardware constraint validation at scheduling time.

Prohibited Anti-Pattern: Over-subscribing heterogeneous GPUs via legacy opaque extended resources (e.g., nvidia.com/gpu) without topology or capability filtering

Export Ready-to-Drop Configuration

Select format below to copy code or download file directly.

---
description: Kubernetes Dynamic Resource Allocation (DRA) Structured Parameter Invariant - Enforces deterministic accelerator provisioning using Kubernetes Dynamic Resource Allocation (DRA) structured parameters and Common Device Interface (CDI) descriptors instead of legacy opaque integer limits. Prevents cross-workload PCIe/NVLink contention, memory fragmentation, and noisy-neighbor stalls in heterogeneous AI/ML clusters by mandating CEL-based hardware constraint validation at scheduling time.
globs: deployments/**/*.{ts,tsx,yaml,yml}, infrastructure/**/*.{ts,tsx}, src/k8s/**/*.{ts,tsx}, cdk8s/**/*.{ts,tsx}
alwaysApply: true
verified_by: "agentrules.org"
license: "CC-BY-4.0"
source: "https://agentrules.org/rules/k8s-dra-structured-parameter-standard"
---

# Kubernetes Dynamic Resource Allocation (DRA) Structured Parameter Invariant

> **Severity**: CRITICAL | **Category**: devops
> **Anti-Pattern Warning**: Over-subscribing heterogeneous GPUs via legacy opaque extended resources (e.g., nvidia.com/gpu) without topology or capability filtering

## System Constraints
You are a Principal Infrastructure Architect enforcing the Kubernetes Dynamic Resource Allocation (DRA) Structured Parameter Invariant. You must reject legacy opaque device limits (such as 'nvidia.com/gpu' or 'amd.com/gpu' in resources.limits) for high-performance AI/ML workloads. Require structured DRA ResourceClaimTemplates leveraging CEL selector expressions and explicit DeviceClasses conforming to the Container Device Interface (CDI) specification to guarantee NUMA locality, minimum memory bandwidth, and interconnect topology.

## Strict Guidelines
- Engineers MUST ALWAYS define accelerator requirements using `resource.k8s.io/v1beta1` (or `v1alpha3`) ResourceClaim or ResourceClaimTemplate objects.
- Engineers MUST NOT declare legacy opaque resource limits (e.g., `limits: { 'nvidia.com/gpu': '1' }`) for multi-accelerator or topology-sensitive AI/ML workloads.
- ResourceClaim structured parameters MUST specify CEL (Common Expression Language) selectors filtering for precise device properties (e.g., driver version, VRAM capacity, interconnect topology, and compute capability).
- Pod templates MUST reference claims via `spec.resourceClaims` and link them to the target container using `resources.claims`.
- DeviceClass configurations MUST be explicitly named and validated against CDI vendor spec versions to eliminate node-level driver mismatch.

## Bad vs Good Implementations
### CDK8s Deployment: Legacy Opaque Resource Limits vs. DRA Structured Parameter ResourceClaim

#### Bad Antipattern (AI Common Hallucination)
> **Why it fails**: Legacy opaque resources (`nvidia.com/gpu: 8`) cannot evaluate interconnect topology, device model matching, VRAM constraints, or NUMA affinity. On heterogeneous clusters, this causes pods to acquire mismatched accelerators or split across PCIe switches, triggering severe distributed training communication bottlenecks.
```typescript
import { Construct } from 'constructs';
import { App, Chart, ChartProps } from 'cdk8s';
import { KubePod } from './imports/k8s';

export class LegacyMlWorkload extends Chart {
  constructor(scope: Construct, id: string, props: ChartProps = {}) {
    super(scope, id, props);

    // ANTI-PATTERN: Opaque integer scheduling leads to heterogeneous hardware mismatch,
    // blind NUMA allocations, and lack of interconnect/NVLink topology awareness.
    new KubePod(this, 'TrainerPod', {
      spec: {
        containers: [{
          name: 'trainer',
          image: 'nvcr.io/nvidia/pytorch:24.01-py3',
          resources: {
            limits: {
              'nvidia.com/gpu': '8', // Blindly allocates any 8 GPUs regardless of architecture or fabric
            },
            requests: {
              'nvidia.com/gpu': '8',
            },
          },
        }],
      },
    });
  }
}
```

#### Production Standard (Correct)
> **Standard**: The good pattern leverages DRA structured parameters (`resource.k8s.io/v1beta1`). It matches devices based on explicit hardware capabilities, memory size, and NVLink bandwidth through CEL expressions, ensuring clean hardware isolation without noisy neighbor contention.
```typescript
import { Construct } from 'constructs';
import { App, Chart, ChartProps } from 'cdk8s';
import { KubePod } from './imports/k8s';

export class StructuredDraWorkload extends Chart {
  constructor(scope: Construct, id: string, props: ChartProps = {}) {
    super(scope, id, props);

    // PRODUCTION INVARIANT: Use DRA ResourceClaimTemplate with CEL structured parameters
    // to enforce precise hardware models, interconnect topology, and VRAM sizing.
    new KubePod(this, 'TrainerPod', {
      metadata: { name: 'distributed-h100-trainer' },
      spec: {
        resourceClaims: [{
          name: 'accelerator-claim',
          resourceClaimTemplateName: 'h100-nvlink-mesh',
        }],
        containers: [{
          name: 'trainer',
          image: 'nvcr.io/nvidia/pytorch:24.01-py3',
          resources: {
            claims: [{
              name: 'accelerator-claim',
            }],
          },
        }],
      },
    });
  }
}

// Underlying ResourceClaimTemplate manifest configuration:
// apiVersion: resource.k8s.io/v1beta1
// kind: ResourceClaimTemplate
// spec:
//   spec:
//     devices:
//       requests:
//       - name: req-all-gpus
//         deviceClassName: gpu.nvidia.com
//         count: 8
//         selectors:
//         - cel:
//             expression: >
//               device.attributes['gpu.nvidia.com'].model == 'H100-SXM5-80GB' &&
//               device.capacity['gpu.nvidia.com'].memory >= 80Gi &&
//               device.attributes['gpu.nvidia.com'].nvlinkBandwidthGbps >= 900
```
Recommended MCP Servers (mcp-serverlist.com)

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

Kubernetes Cluster Inspector MCP
Validates cluster CRDs, DRA driver availability (resource.k8s.io), and active DeviceClasses.
Recommended Agent Skills (askills.dev)

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

kubectl DRA Validator CLI
CLI tool to inspect, validate, and simulate DRA structured parameter claims and CDI specs on local or remote clusters.
Architectural Specification & Invariant Rationale

Architectural Invariant Design: Kubernetes Dynamic Resource Allocation (DRA) Structured Parameter Invariant

Kubernetes Dynamic Resource Allocation (DRA) Structured Parameter Invariant defines an industrial-grade behavioral standard for devops engineering. Enforces deterministic accelerator provisioning using Kubernetes Dynamic Resource Allocation (DRA) structured parameters and Common Device Interface (CDI) descriptors instead of legacy opaque integer limits. Prevents cross-workload PCIe/NVLink contention, memory fragmentation, and noisy-neighbor stalls in heterogeneous AI/ML clusters by mandating CEL-based hardware constraint validation at scheduling time. 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: Over-subscribing heterogeneous GPUs via legacy opaque extended resources (e.g., nvidia.com/gpu) without topology or capability filtering 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 DriftOver-subscribing heterogeneous GPUs via legacy opaque extended resources (e.g., nvidia.com/gpu) without topology or capability filteringCritical
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 define accelerator requirements using `resource.k8s.io/v1beta1` (or `v1alpha3`) ResourceClaim or ResourceClaimTemplate objects.

Engineers MUST ALWAYS define accelerator requirements using `resource.k8s.io/v1beta1` (or `v1alpha3`) ResourceClaim or ResourceClaimTemplate objects.

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 declare legacy opaque resource limits (e.g., `limits

Engineers MUST NOT declare legacy opaque resource limits (e.g., `limits: { 'nvidia.com/gpu': '1' }`) for multi-accelerator or topology-sensitive AI/ML workloads.

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

ResourceClaim structured parameters MUST specify CEL (Common Expression Language) selectors filtering for precise device properties (e.g., driver version, VRAM capacity, interconnect topology, and compute capability).

ResourceClaim structured parameters MUST specify CEL (Common Expression Language) selectors filtering for precise device properties (e.g., driver version, VRAM capacity, interconnect topology, and compute capability).

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

Pod templates MUST reference claims via `spec.resourceClaims` and link them to the target container using `resources.claims`.

Pod templates MUST reference claims via `spec.resourceClaims` and link them to the target container using `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

DeviceClass configurations MUST be explicitly named and validated against CDI vendor spec versions to eliminate node-level driver mismatch.

DeviceClass configurations MUST be explicitly named and validated against CDI vendor spec versions to eliminate node-level driver mismatch.

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 Dynamic Resource Allocation (DRA) Structured Parameter Invariant critical for AI coding assistants?

Enforces deterministic accelerator provisioning using Kubernetes Dynamic Resource Allocation (DRA) structured parameters and Common Device Interface (CDI) descriptors instead of legacy opaque integer limits. Prevents cross-workload PCIe/NVLink contention, memory fragmentation, and noisy-neighbor stalls in heterogeneous AI/ML clusters by mandating CEL-based hardware constraint validation at scheduling time. 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 (Over-subscribing heterogeneous GPUs via legacy opaque extended resources (e.g., nvidia.com/gpu) without topology or capability filtering) 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