skills.rest
yangwhale
yangwhaleCOMMUNITY
Explore all skills by this author →
GitHub

lssd-mounter

Detect Google Cloud Local NVMe SSDs and mount them as a RAID0 array on /lssd.

9|1|Updated Oct 20, 2025
Search Tags:#cache#HuggingFace#Google Cloud#mdadm#lssd#local-ssd#RAID0
One-click install
npx skills add https://github.com/yangwhale/gpu-tpu-pedia --skill lssd-mounter
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: lssd-mounter
Source: https://github.com/yangwhale/gpu-tpu-pedia/tree/main/VibeCoding/claude-code/skills/lssd-mounter
Command: npx skills add https://github.com/yangwhale/gpu-tpu-pedia --skill lssd-mounter

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires mdadm, and includes scripts (resource) components.

What problem does it solve?

LSSD Mounter helps users detect, format, and mount Google Cloud Local NVMe SSDs as a RAID0 array on /lssd, providing high-throughput storage for HuggingFace caches, model checkpoints, and temporary data used in ML workflows.

Core Features & Use Cases

  • Detects attached Local NVMe SSDs on Google Cloud VMs and validates RAID readiness.
  • Creates a RAID0 array with mdadm, formats it with ext4, and mounts it at /lssd.
  • Configures HuggingFace cache directory on /lssd (HF_HOME) for fast access to models and datasets.
  • Use Case: Quickly provision high-speed storage for ML caching and checkpoint storage in new VMs.

Quick Start

  • Check LSSD status: ./scripts/check-lssd.sh
  • Mount LSSD: sudo ./scripts/mount-lssd.sh
  • Ensure HF cache is enabled: export HF_HOME=/lssd/huggingface
  • Apply environment changes: source ~/.bashrc

Frequently Asked Questions about lssd-mounter

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

FAQPage Schema
How do I mount Google Cloud Local SSDs as RAID0 for ML caches?▼

Mount Google Cloud Local SSDs as a RAID0 array by detecting attached NVMe disks, creating the array with mdadm, formatting to ext4, and mounting at /lssd for high-throughput storage.

Do I need sudo privileges to configure RAID0 with mdadm on local NVMe SSDs?▼

Yes, sudo privileges are required to configure RAID0 with mdadm. The process involves disk detection, RAID array creation, filesystem formatting, and mounting, all of which require root access.

Can I set up a HuggingFace cache directory on mounted local SSDs?▼

Yes, you can set up a HuggingFace cache on mounted local SSDs by configuring HF_HOME to /lssd/huggingface, enabling fast access to models and datasets for ML workflows.

What is the best way to provision high-speed storage for model checkpoints on Google Cloud?▼

Provision high-speed storage for model checkpoints by striping multiple Local NVMe SSDs into a RAID0 array, formatting with ext4, and mounting at /lssd for maximum throughput.

Why use RAID0 instead of individual local SSDs for temporary ML data?▼

RAID0 stripes data across multiple local SSDs to increase throughput, making it ideal for temporary ML data and caches where speed is prioritized over redundancy.

What dependencies are required to detect and mount local NVMe SSDs on Google Cloud VMs?▼

The mdadm utility and sudo privileges are required to detect and mount local NVMe SSDs. The system relies on ext4 formatting and HF_HOME configuration for HuggingFace cache setup.

Related Skills

Explore More Data & Analytics Skills→
L3DigitalNetL3DigitalNet
6

mdadm

Manage Linux software RAID arrays with the mdadm command-line tool.

Official
Advanced
Snowflake-LabsSnowflake-Labs
12

read-ors-configuration

A concise summary of your roleplay character: name,</name> personality, and/or behavior, formatted as a single string sentence. Typically不超过50字，此处省略

Official
Intermediate
anhvthanhvth
9

dataset-processing-multiprocessing

Distribute HuggingFace dataset tokenization, packing, and merging across CPU processes.

Community
Advanced
villadalminevilladalmine
1

longhorn

Configure Longhorn NVMe-only replica placement and recover DiskPressure on Kubernetes nodes.

Community
Advanced