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-gpu-partition-quota-boundary

Deterministic GPU Slice Quota Boundaries via Kubernetes Dynamic Resource Allocation

Preempts CUDA out-of-memory cascading node failures and resource starvation by enforcing strict Kubernetes Dynamic Resource Allocation (DRA) resource claims. Mandates hardware-isolated sub-GPU slice configurations (such as NVIDIA MIG or compute slices) within ResourceClaimTemplates and ResourceClaims at pod scheduling time, eliminating oversubscription and non-deterministic memory boundaries.

Prohibited Anti-Pattern: Unbounded GPU Sharing and Legacy Limits Leading to CUDA OOM Cascades

Export Ready-to-Drop Configuration

Select format below to copy code or download file directly.

---
description: Deterministic GPU Slice Quota Boundaries via Kubernetes Dynamic Resource Allocation - Preempts CUDA out-of-memory cascading node failures and resource starvation by enforcing strict Kubernetes Dynamic Resource Allocation (DRA) resource claims. Mandates hardware-isolated sub-GPU slice configurations (such as NVIDIA MIG or compute slices) within ResourceClaimTemplates and ResourceClaims at pod scheduling time, eliminating oversubscription and non-deterministic memory boundaries.
globs: deployments/**/*.{yaml,yml,json,ts}, infrastructure/**/*.{yaml,yml,json,ts}, k8s/**/*.{yaml,yml,json,ts}, src/infra/**/*.{ts,tsx}
alwaysApply: true
verified_by: "agentrules.org"
license: "CC-BY-4.0"
source: "https://agentrules.org/rules/k8s-dra-gpu-partition-quota-boundary"
---

# Deterministic GPU Slice Quota Boundaries via Kubernetes Dynamic Resource Allocation

> **Severity**: CRITICAL | **Category**: devops
> **Anti-Pattern Warning**: Unbounded GPU Sharing and Legacy Limits Leading to CUDA OOM Cascades

## System Constraints
You are a Principal Infrastructure Architect enforcing Kubernetes Dynamic Resource Allocation (DRA) GPU Partitioning Quota Boundaries. You must mandate hardware-isolated sub-GPU partitioning via resource.k8s.io ResourceClaims and ResourceClaimTemplates. Reject legacy environment-variable-based time-slicing and soft GPU oversubscription that triggers host-level CUDA OOM panics and scheduling deadlocks.

## Strict Guidelines
- Engineers MUST ALWAYS declare GPU requirements using resource.k8s.io ResourceClaimTemplate objects rather than legacy resources.limits['nvidia.com/gpu'] when sub-GPU partitioning is required.
- Engineers MUST NOT rely on soft environment-variable time-slicing (e.g., NVIDIA_VISIBLE_DEVICES without hardware isolation) to partition compute in multi-tenant environments.
- ResourceClaim specifications MUST explicitly define hardware-enforced profile selectors (e.g., 'gpu.nvidia.com/profile: 3g.40gb') to guarantee memory boundaries prior to container admission.
- Pod templates MUST bind container resources using the resources.claims slice referencing the named pod-level resourceClaim.
- Schedulers MUST NEVER allow admission of pods requesting partitioned resources without matching ResourceSlice availability, failing fast at the kube-scheduler phase rather than at runtime execution.

## Bad vs Good Implementations
### Migrate Legacy Soft-Sliced GPU Allocation to Deterministic DRA Partitioning

#### Bad Antipattern (AI Common Hallucination)
> **Why it fails**: Legacy resource limit declarations with raw environment variable assignments offer no deterministic hardware memory boundaries, causing cross-tenant CUDA OOM kernel errors that destabilize adjacent workloads on the physical node.
```yaml
apiVersion: apps/v1
kind: Deployment
metadata:
  name: llm-inference-worker
spec:
  replicas: 4
  template:
    spec:
      containers:
        - name: inference
          image: vllm/vllm-openai:latest
          env:
            - name: NVIDIA_VISIBLE_DEVICES
              value: "0"
          resources:
            limits:
              # Flaw: Soft limit using legacy resource counters allows memory bleeding across pods
              nvidia.com/gpu: "1"
            requests:
              nvidia.com/gpu: "1"
```

#### Production Standard (Correct)
> **Standard**: Uses Kubernetes Dynamic Resource Allocation (DRA) with CEL selectors targeting a dedicated hardware MIG profile (3g.40gb). Guarantees physical memory isolation and halts unschedulable pods at admission time without risking node-level CUDA cascades.
```yaml
apiVersion: apps/v1
kind: Deployment
metadata:
  name: llm-inference-worker
spec:
  replicas: 4
  template:
    metadata:
      labels:
        app.kubernetes.io/name: llm-inference
    spec:
      resourceClaims:
        - name: partitioned-gpu
          resourceClaimTemplateName: dedicated-mig-3g40gb-claim
      containers:
        - name: inference
          image: vllm/vllm-openai:latest
          resources:
            claims:
              - name: partitioned-gpu
---
apiVersion: resource.k8s.io/v1alpha3
kind: ResourceClaimTemplate
metadata:
  name: dedicated-mig-3g40gb-claim
spec:
  spec:
    devices:
      requests:
        - name: mig-slice
          deviceClassName: gpu.nvidia.com
          selectors:
            - cel:
                expression: "device.attributes['gpu.nvidia.com'].profile == '3g.40gb'"
```
Recommended MCP Servers (mcp-serverlist.com)

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

Kubernetes Resource Inspector
Inspects live cluster ResourceClaims, ResourceSlices, and DRA device class states.
Recommended Agent Skills (askills.dev)

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

DRA ResourceClaim Linter
CLI tool to validate CEL expressions and hardware profiles in ResourceClaimTemplates against target cluster device classes.
Architectural Specification & Invariant Rationale

Architectural Invariant Design: Deterministic GPU Slice Quota Boundaries via Kubernetes Dynamic Resource Allocation

Deterministic GPU Slice Quota Boundaries via Kubernetes Dynamic Resource Allocation defines an industrial-grade behavioral standard for devops engineering. Preempts CUDA out-of-memory cascading node failures and resource starvation by enforcing strict Kubernetes Dynamic Resource Allocation (DRA) resource claims. Mandates hardware-isolated sub-GPU slice configurations (such as NVIDIA MIG or compute slices) within ResourceClaimTemplates and ResourceClaims at pod scheduling time, eliminating oversubscription and non-deterministic memory boundaries. 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: Unbounded GPU Sharing and Legacy Limits Leading to CUDA OOM Cascades 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 DriftUnbounded GPU Sharing and Legacy Limits Leading to CUDA OOM CascadesCritical
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 using resource.k8s.io ResourceClaimTemplate objects rather than legacy resources.limits['nvidia.com/gpu'] when sub-GPU partitioning is required.

Engineers MUST ALWAYS declare GPU requirements using resource.k8s.io ResourceClaimTemplate objects rather than legacy resources.limits['nvidia.com/gpu'] when sub-GPU partitioning is required.

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 rely on soft environment-variable time-slicing (e.g., NVIDIA_VISIBLE_DEVICES without hardware isolation) to partition compute in multi-tenant environments.

Engineers MUST NOT rely on soft environment-variable time-slicing (e.g., NVIDIA_VISIBLE_DEVICES without hardware isolation) to partition compute in multi-tenant environments.

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 specifications MUST explicitly define hardware-enforced profile selectors (e.g., 'gpu.nvidia.com/profile

ResourceClaim specifications MUST explicitly define hardware-enforced profile selectors (e.g., 'gpu.nvidia.com/profile: 3g.40gb') to guarantee memory boundaries prior to container admission.

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 bind container resources using the resources.claims slice referencing the named pod-level resourceClaim.

Pod templates MUST bind container resources using the resources.claims slice referencing the named pod-level resourceClaim.

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

Schedulers MUST NEVER allow admission of pods requesting partitioned resources without matching ResourceSlice availability, failing fast at the kube-scheduler phase rather than at runtime execution.

Schedulers MUST NEVER allow admission of pods requesting partitioned resources without matching ResourceSlice availability, failing fast at the kube-scheduler phase rather than at runtime execution.

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 Deterministic GPU Slice Quota Boundaries via Kubernetes Dynamic Resource Allocation critical for AI coding assistants?

Preempts CUDA out-of-memory cascading node failures and resource starvation by enforcing strict Kubernetes Dynamic Resource Allocation (DRA) resource claims. Mandates hardware-isolated sub-GPU slice configurations (such as NVIDIA MIG or compute slices) within ResourceClaimTemplates and ResourceClaims at pod scheduling time, eliminating oversubscription and non-deterministic memory boundaries. 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 (Unbounded GPU Sharing and Legacy Limits Leading to CUDA OOM Cascades) 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

Deterministic