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-device-partitioning-invariant

Kubernetes 1.32 DRA Structured Device Partitioning and Tenant Isolation

Enforces deterministic, conflict-free scheduling and strict hardware slice isolation for heterogeneous GPU and NPU inference workloads using Kubernetes 1.32 Dynamic Resource Allocation (DRA) with resource.k8s.io/v1beta1. Prevents co-location deadlocks and multi-tenant memory leakage by mandating structured parameter selectors, explicit partitioning attributes, and disallowing legacy Opaque Integer Resource limits alongside DRA claims.

Prohibited Anti-Pattern: Mixing legacy resource limits with DRA claims and omitting CEL device partition constraints

Export Ready-to-Drop Configuration

Select format below to copy code or download file directly.

---
description: Kubernetes 1.32 DRA Structured Device Partitioning and Tenant Isolation - Enforces deterministic, conflict-free scheduling and strict hardware slice isolation for heterogeneous GPU and NPU inference workloads using Kubernetes 1.32 Dynamic Resource Allocation (DRA) with resource.k8s.io/v1beta1. Prevents co-location deadlocks and multi-tenant memory leakage by mandating structured parameter selectors, explicit partitioning attributes, and disallowing legacy Opaque Integer Resource limits alongside DRA claims.
globs: deploy/**/*.yaml, k8s/**/*.yaml, helm/**/*.yaml, src/infrastructure/**/*.ts, src/controllers/**/*.ts
alwaysApply: true
verified_by: "agentrules.org"
license: "CC-BY-4.0"
source: "https://agentrules.org/rules/k8s-132-dra-device-partitioning-invariant"
---

# Kubernetes 1.32 DRA Structured Device Partitioning and Tenant Isolation

> **Severity**: CRITICAL | **Category**: devops
> **Anti-Pattern Warning**: Mixing legacy resource limits with DRA claims and omitting CEL device partition constraints

## System Constraints
You are a Principal DevOps and Infrastructure Architect enforcing Kubernetes 1.32 Dynamic Resource Allocation (DRA) standards for heterogeneous accelerator fleets. You MUST mandate resource.k8s.io/v1beta1 constructs, explicit device partitioning, CEL selector attributes, and strict isolation invariants across all workloads.

## Strict Guidelines
- Engineers MUST ALWAYS declare accelerator requirements using 'resource.k8s.io/v1beta1' ResourceClaim or ResourceClaimTemplate objects instead of legacy 'resources.limits["nvidia.com/gpu"]' or vendor-specific Opaque Integer Resources.
- Engineers MUST NOT mix legacy resource limits and DRA claims in the same Pod specification to avoid kube-scheduler two-phase allocation deadlocks.
- Services MUST enforce Common Expression Language (CEL) selectors in 'DeviceRequest.selectors' to specify exact compute capability, partition profile (e.g., MIG profile, NPU core count), and memory minimums.
- Workloads requiring hardware partitioning MUST explicitly define 'allocationMode: ExactCount' with allocation constraints to ensure device slice boundary adherence.
- Components NEVER permit 'adminAccess: true' on ResourceClaims deployed in multi-tenant namespaces; elevated hardware access MUST remain restricted to platform control plane tooling.

## Bad vs Good Implementations
### Kubernetes 1.32 ResourceClaim DRA Manifest Definition (TypeScript/cdk8s)

#### Bad Antipattern (AI Common Hallucination)
> **Why it fails**: Legacy resource limit slicing relies on kubelet-level opaque integers without structured scheduler visibility. It permits co-location of mismatched slice topologies, fails to guarantee memory/bandwidth isolation across tenants, and cannot evaluate runtime hardware constraints.
```typescript
import { KubePod } from 'cdk8s-plus-31';

// ANTI-PATTERN: Relying on legacy opaque integer limits for heterogeneous partitions
// Causes noisy-neighbor collision, lacks slice isolation, and risks scheduling race conditions.
export function createInferencePod(scope: any, id: string) {
  return {
    apiVersion: 'v1',
    kind: 'Pod',
    metadata: { name: 'llm-inference-worker', namespace: 'tenant-a' },
    spec: {
      containers: [{
        name: 'engine',
        image: 'vllm/vllm-openai:v0.6.0',
        resources: {
          limits: {
            'nvidia.com/mig-3g.40gb': '1', // Legacy unstructured slicing
            'cpu': '8',
            'memory': '32Gi'
          }
        }
      }]
    }
  };
}
```

#### Production Standard (Correct)
> **Standard**: Leverages Kubernetes 1.32 DRA (resource.k8s.io/v1beta1) with structured parameters. Scheduler evaluates CEL expressions to enforce deterministic slice isolation, guarantees exact-match hardware boundaries, and eliminates multi-tenant noisy neighbor cross-talk.
```typescript
import { Construct } from 'constructs';

// PRODUCTION INVARIANT: K8s 1.32 DRA resource.k8s.io/v1beta1 structured parameter specification
export function createDraInferencePod(scope: Construct, id: string) {
  return {
    apiVersion: 'v1',
    kind: 'Pod',
    metadata: { name: 'llm-inference-worker', namespace: 'tenant-a' },
    spec: {
      resourceClaims: [
        { name: 'gpu-partition', resourceClaimTemplateName: 'gpu-partition-template' }
      ],
      containers: [{
        name: 'engine',
        image: 'vllm/vllm-openai:v0.6.0',
        resources: {
          claims: [{ name: 'gpu-partition' }],
          limits: { cpu: '8', memory: '32Gi' }
        }
      }]
    }
  };
}

export function createDraClaimTemplate(scope: Construct, id: string) {
  return {
    apiVersion: 'resource.k8s.io/v1beta1',
    kind: 'ResourceClaimTemplate',
    metadata: { name: 'gpu-partition-template', namespace: 'tenant-a' },
    spec: {
      spec: {
        devices: {
          requests: [{
            name: 'mig-slice',
            deviceClassName: 'gpu.nvidia.com',
            allocationMode: 'ExactCount',
            count: 1,
            adminAccess: false,
            selectors: [{
              cel: {
                expression: "device.attributes['gpu.nvidia.com'].sliceSize == '3g.40gb' && device.attributes['gpu.nvidia.com'].family == 'hopper'"
              }
            }]
          }]
        }
      }
    }
  };
}
```
Recommended MCP Servers (mcp-serverlist.com)

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

Kubernetes Cluster Inspector MCP
Queries live cluster API definitions to validate DRA driver custom resource definitions and device class availability.
Recommended Agent Skills (askills.dev)

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

DRA CEL and Resource Claim Syntax Validator
CLI utility to lint and dry-run resource.k8s.io/v1beta1 ResourceClaim specs and CEL syntax against Kubernetes 1.32 API schemas.
Architectural Specification & Invariant Rationale

Architectural Invariant Design: Kubernetes 1.32 DRA Structured Device Partitioning and Tenant Isolation

Kubernetes 1.32 DRA Structured Device Partitioning and Tenant Isolation defines an industrial-grade behavioral standard for devops engineering. Enforces deterministic, conflict-free scheduling and strict hardware slice isolation for heterogeneous GPU and NPU inference workloads using Kubernetes 1.32 Dynamic Resource Allocation (DRA) with resource.k8s.io/v1beta1. Prevents co-location deadlocks and multi-tenant memory leakage by mandating structured parameter selectors, explicit partitioning attributes, and disallowing legacy Opaque Integer Resource limits alongside DRA claims. 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: Mixing legacy resource limits with DRA claims and omitting CEL device partition constraints 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 DriftMixing legacy resource limits with DRA claims and omitting CEL device partition constraintsCritical
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 accelerator requirements using 'resource.k8s.io/v1beta1' ResourceClaim or ResourceClaimTemplate objects instead of legacy 'resources.limits["nvidia.com/gpu"]' or vendor-specific Opaque Integer Resources.

Engineers MUST ALWAYS declare accelerator requirements using 'resource.k8s.io/v1beta1' ResourceClaim or ResourceClaimTemplate objects instead of legacy 'resources.limits["nvidia.com/gpu"]' or vendor-specific Opaque Integer Resources.

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 mix legacy resource limits and DRA claims in the same Pod specification to avoid kube-scheduler two-phase allocation deadlocks.

Engineers MUST NOT mix legacy resource limits and DRA claims in the same Pod specification to avoid kube-scheduler two-phase allocation deadlocks.

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

Services MUST enforce Common Expression Language (CEL) selectors in 'DeviceRequest.selectors' to specify exact compute capability, partition profile (e.g., MIG profile, NPU core count), and memory minimums.

Services MUST enforce Common Expression Language (CEL) selectors in 'DeviceRequest.selectors' to specify exact compute capability, partition profile (e.g., MIG profile, NPU core count), and memory minimums.

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

Workloads requiring hardware partitioning MUST explicitly define 'allocationMode

Workloads requiring hardware partitioning MUST explicitly define 'allocationMode: ExactCount' with allocation constraints to ensure device slice boundary adherence.

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

Components NEVER permit 'adminAccess

Components NEVER permit 'adminAccess: true' on ResourceClaims deployed in multi-tenant namespaces; elevated hardware access MUST remain restricted to platform control plane tooling.

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 Device Partitioning and Tenant Isolation critical for AI coding assistants?

Enforces deterministic, conflict-free scheduling and strict hardware slice isolation for heterogeneous GPU and NPU inference workloads using Kubernetes 1.32 Dynamic Resource Allocation (DRA) with resource.k8s.io/v1beta1. Prevents co-location deadlocks and multi-tenant memory leakage by mandating structured parameter selectors, explicit partitioning attributes, and disallowing legacy Opaque Integer Resource limits alongside DRA claims. 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 (Mixing legacy resource limits with DRA claims and omitting CEL device partition constraints) 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