skills.rest
sayalinvidia
sayalinvidiaCOMMUNITY
Explore all skills by this author →
GitHub

dynamo-interconnect-check

Assess NIXL/UCX/NCCL transport readiness for Dynamo disaggregated serving.

Updated Apr 16, 2026
Search Tags:#verification#read-only#interconnect#nccl#dynamo#nixl#ucx
One-click install
npx skills add https://github.com/sayalinvidia/sayali-skills-test --skill dynamo-interconnect-check
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: dynamo-interconnect-check
Source: https://github.com/sayalinvidia/sayali-skills-test/tree/main/plugins/nvidia-skills/skills/dynamo-interconnect-check
Command: npx skills add https://github.com/sayalinvidia/sayali-skills-test --skill dynamo-interconnect-check

SYSTEM DOCUMENTATION & REQUIREMENTS

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

What problem does it solve?

Interconnect readiness between Dynamo's prefill and decode workers over NIXL/UCX/NCCL can silently be wrong even when a deployment seems healthy. This skill validates the transport readiness to ensure fast, correct KV transfers for disaggregated serving.

Core Features & Use Cases

  • Read-only checks of transport-related environment variables on deployment manifests.
  • Runtime capability probes for InfiniBand/GPUDirect/NVLink on target nodes or pods.
  • Confirm NIXL availability or provide next steps when tools are missing, guiding safe deployment decisions.
  • Use after dynamo recipe deployment to verify transport health before benchmarking.

Quick Start

Run the read-only interconnect checks after deploying a Dynamo recipe to validate KV transport readiness before benchmarking.

Frequently Asked Questions about dynamo-interconnect-check

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

FAQPage Schema
How do I verify NIXL and UCX transport readiness for Dynamo disaggregated serving?▼

Verify NIXL and UCX transport readiness by running read-only checks against deployment manifests and probing node capabilities. This validates that KV transport uses the fast path and prevents silent performance degradation.

Why does NCCL interconnect performance silently degrade in Dynamo recipes?▼

NCCL interconnect performance degrades silently when transport environment variables are misconfigured in deployment manifests. Interconnect readiness checks detect these mismatches by probing node capabilities and verifying NIXL availability.

How do I check InfiniBand and GPUDirect availability in Kubernetes pods for Dynamo?▼

Check InfiniBand and GPUDirect availability by probing target pods and nodes in read-only mode. This confirms whether the necessary hardware capabilities exist for fast KV transfers without exposing secrets.

When do I need to validate NIXL tooling availability before benchmarking Dynamo?▼

Validate NIXL tooling availability immediately after deploying a Dynamo recipe and before benchmarking. This ensures the transport layer is healthy and guides safe deployment decisions if required tools are missing.

Can I inspect transport environment variables in deploy manifests without exposing secrets?▼

Inspect transport environment variables in deploy manifests using read-only checks that never expose secrets. This safely assesses NIXL, UCX, and NCCL configuration to ensure correct KV transfers.

Related Skills

Explore More Software Engineering Skills→
sayalinvidiasayalinvidia

dynamo-recipe-runner

Validate and patch existing Dynamo Kubernetes recipes, then deploy and smoke-test endpoints.

Community
Advanced
InverterNetworkInverterNetwork

tensorrt-llm

Optimize LLM inference on NVIDIA GPUs with quantization and parallelization.

Official
Advanced
op7418op7418
136

card-wifi-setup

Provision Wi-Fi on the AI Desk Card via Bash and AskUserQuestion.

Community
Intermediate
magnus919magnus919
40

headscale-derp

Deploy and manage DERP relay servers for Tailscale and Headscale networks.

Community
Advanced