skills.rest
villadalmine
villadalmineCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

longhorn

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

1|Updated Apr 3, 2026
Search Tags:#kubernetes#nvme#storageclass#longhorn#replicas#emmc#diskpressure
One-click install
npx skills add https://github.com/villadalmine/infra --skill longhorn
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: longhorn
Source: https://github.com/villadalmine/infra/tree/main/.agents/skills/longhorn
Command: npx skills add https://github.com/villadalmine/infra --skill longhorn

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill prevents Longhorn from wasting replicas on tiny eMMC disks and causing DiskPressure, pod evictions, and storage-related outages in mixed-disk Kubernetes clusters.

Core Features & Use Cases

  • NVMe-First Storage Design: Guides Longhorn to place replicas on NVMe-backed disks instead of the default eMMC path.
  • StorageClass and Node Policy Tuning: Helps configure disk selectors, node annotations, and default disk settings so scheduling stays predictable.
  • DiskPressure Recovery: Supports emergency remediation when replicas are already trapped on eMMC and nodes need eviction or rescheduling.
  • Use Case: An RK1 node hosting OpenClaw or NPU workloads starts filling its root disk; this Skill helps move storage to NVMe before Kubernetes begins evicting critical pods.

Quick Start

Use the longhorn skill to review the cluster’s disk layout, verify NVMe-only replica placement, and recommend the Longhorn and StorageClass changes needed to prevent DiskPressure.

Frequently Asked Questions about longhorn

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

FAQPage Schema
Why does Longhorn storage cause DiskPressure on my Kubernetes nodes?▼

Longhorn causes DiskPressure when it schedules replicas onto small eMMC disks instead of NVMe, filling the root disk and triggering pod evictions. Applying NVMe-only disk selectors prevents this storage issue.

How do I force Longhorn to use NVMe disks instead of eMMC?▼

You force Longhorn to use NVMe by applying StorageClass disk selectors and node disk annotations. This enforces NVMe-only scheduling for replicas, bypassing the default eMMC path on mixed-disk nodes.

Can I evict Longhorn replicas from an eMMC disk after it fills up?▼

Yes, you can evict Longhorn replicas from eMMC using emergency API patches. This remediates DiskPressure by moving trapped replicas to NVMe disks, rescheduling critical workloads without data loss.

Does this Longhorn configuration work for mixed eMMC and NVMe RK1 nodes?▼

Yes, this configuration targets RK1 nodes with mixed eMMC and NVMe disks. It tunes Longhorn node annotations and default disk settings to ensure predictable NVMe storage scheduling for workloads like OpenClaw and NPU services.

What is the best way to configure StorageClass for Longhorn block storage?▼

The best way to configure Longhorn StorageClass is by applying disk selectors that target NVMe paths. This ensures predictable replica placement and prevents storage-related outages on mixed-disk Kubernetes clusters.

Related Skills

Explore More Software Engineering Skills→
tomazbtomazb
2

openshift-longhorn

Automate Longhorn lifecycle operations on OpenShift/OKD clusters.

Community
Advanced
wcyganwcygan
3

longhorn-volume-ops

Create, resize, clone, and tune per-volume Longhorn PVCs in Anton.

Community
Intermediate
Probably-GroupProbably-Group
3

talos-cluster-ops

Correlate Talos, Kubernetes, Cilium, Longhorn, and GitOps evidence to diagnose cluster failures.

Official
Advanced
wcyganwcygan
3

longhorn-node-ops

Align Longhorn disk configurations with Talos per-node patches via kubectl.

Community
Advanced