devops:gpu-analysis

Analyze IDC GPU cluster health and VRAM usage via kubectl and Grafana DCGM metrics.

Updated Feb 27, 2026

Search Tags:#monitoring #kubernetes #gpu #grafana #vram #resource analysis #dcgm

Ready to run

npx skills add https://github.com/ChanghwanK/dotfiles --skill devops-gpu-analysis

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: devops:gpu-analysis
Source: https://github.com/ChanghwanK/dotfiles/tree/main/claude/.claude/skills/devops%3Agpu-analysis
Command: npx skills add https://github.com/ChanghwanK/dotfiles --skill devops-gpu-analysis

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

This Skill provides a comprehensive analysis of GPU cluster usage and health, helping you understand resource allocation, identify bottlenecks, and determine model deployment feasibility.

Core Features & Use Cases

Cluster Overview: Get a real-time snapshot of GPU node status, VRAM utilization, power consumption, and temperature.
Resource Allocation: Analyze how workloads are mapped to GPUs and identify underutilized or overutilized resources.
Deployment Feasibility: Assess if your models can be deployed based on VRAM availability and requirements.
Troubleshooting: Detect nodes with high load or abnormal conditions.

Quick Start

Analyze the current GPU cluster status and VRAM usage for the k8s-idc context.

Frequently Asked Questions about devops:gpu-analysis

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

FAQPage Schema

How do I monitor GPU cluster health and VRAM usage in Kubernetes?▼

Analyze GPU workload mapping by using kubectl to retrieve node and pod information, then cross-reference with Grafana DCGM metrics. This identifies how workloads are mapped to GPUs and highlights underutilized or overutilized resources.

Can I check VRAM availability to determine model deployment feasibility?▼

Check VRAM availability to determine model deployment feasibility by analyzing current cluster metrics. The Skill assesses whether your models can be deployed based on real-time VRAM availability and specific model requirements across the GPU nodes.

What is the best way to detect abnormal GPU nodes and high load conditions?▼

Detect abnormal GPU nodes and high load conditions by analyzing Grafana DCGM metrics for power, temperature, and VRAM utilization. This troubleshooting capability identifies nodes operating under abnormal conditions or experiencing excessive load.

Do I need Prometheus and Grafana DCGM metrics to analyze IDC GPU resource allocation?▼

You need Prometheus queried via Grafana for DCGM metrics and kubectl for Kubernetes node information to analyze IDC GPU resource allocation. These dependencies are required to retrieve the necessary health, VRAM, and power data for analysis.