aws-fsx

Explain FSx for Lustre architecture and deployment on Kubernetes with S3.

Updated Aug 27, 2026
One-click install
npx skills add https://github.com/tylertitsworth/skills --skill aws-fsx
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: aws-fsx
Source: https://github.com/tylertitsworth/skills/tree/main/aws-fsx
Command: npx skills add https://github.com/tylertitsworth/skills --skill aws-fsx

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

FSx for Lustre on Kubernetes is a complex, high-throughput shared storage solution. This guide explains architecture, deployment patterns, and integration points to help ML workflows run reliably at scale.

Core Features & Use Cases

  • Understand MDT/OST architecture and how to optimize striping, throughput, and metadata IOPS for ML training.
  • Configure Data Repository Associations (DRA) to link S3 data to Lustre filesystems and enable lazy loading.
  • Deploy FSx for Lustre on EKS using the CSI driver, manage encryption, backups, and GPUDirect Storage (GDS) support.
  • Troubleshoot common performance and mount issues on Kubernetes clusters.

Quick Start

Mount FSx for Lustre on your EKS cluster and run a small test workload to validate throughput.

Frequently Asked Questions about aws-fsx

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I configure FSx for Lustre on EKS for ML training workloads?

To configure FSx for Lustre on EKS for ML training, deploy the FSx CSI driver to mount the filesystem, then optimize MDT/OST striping and throughput to handle high-volume checkpointing and data preprocessing operations.

What is Data Repository Association in FSx Lustre and how does it link S3 data?

Data Repository Associations (DRA) link S3 data repositories to FSx Lustre filesystems, enabling lazy loading so ML training jobs can seamlessly access S3 datasets without manual synchronization steps.

Can I use GPUDirect Storage with FSx for Lustre on Kubernetes?

Yes, FSx for Lustre supports GPUDirect Storage (GDS) on Kubernetes to enable direct data access from GPU memory, bypassing the CPU and optimizing throughput for demanding ML training workloads.

What is the best way to tune FSx Lustre striping and metadata IOPS for ML workloads?

The best way to tune FSx Lustre for ML workloads involves adjusting the MDT/OST architecture configuration to optimize striping patterns and metadata IOPS, ensuring high-throughput access for training and checkpointing operations.

Why does my FSx for Lustre mount fail or perform poorly on Kubernetes?

FSx for Lustre mount failures or poor performance on Kubernetes often stem from incorrect CSI driver integration, misconfigured DRA settings, or suboptimal striping configurations that bottleneck metadata IOPS during ML training.