Implementation:LMCache LMCache Disagg Prerequisite Checks

Knowledge Sources	LMCache
Domains	Infrastructure, Deployment
Last Updated	2026-02-09 00:00 GMT

Overview

Concrete tool for validating disaggregated prefill deployment prerequisites provided by the LMCache example scripts.

Description

The disagg_example_1p1d.sh script performs prerequisite checks including GPU count verification via nvidia-smi, Python package import tests, and HuggingFace token validation. This is a Pattern Doc documenting the expected validation interface.

Usage

Run the orchestration script before deploying disaggregated prefill. It will abort with clear error messages if prerequisites are not met.

Code Reference

Source Location

• Repository: LMCache
• File: examples/disagg_prefill/1p1d/disagg_example_1p1d.sh
• Lines: L1-L50

Signature

#!/bin/bash
# Prerequisite checks:
# 1. Check GPU count >= 2
# 2. Check Python packages: lmcache, vllm, nixl, zmq, httpx
# 3. Check HF_TOKEN environment variable
# 4. Set MODEL variable (default: meta-llama/Llama-3.1-8B-Instruct)

Import

# Run directly as shell script
bash examples/disagg_prefill/1p1d/disagg_example_1p1d.sh

I/O Contract

Inputs

Name	Type	Required	Description
HF_TOKEN	env var	Yes	HuggingFace authentication token
NVIDIA GPUs	hardware	Yes	Minimum 2 GPUs for 1P1D deployment
Python packages	software	Yes	lmcache, vllm, nixl, zmq, httpx must be importable

Outputs

Name	Type	Description
validation result	exit code	0 if all checks pass, non-zero on failure

Usage Examples

Manual Prerequisite Check

# Check GPU count
GPU_COUNT=$(nvidia-smi --query-gpu=name --format=csv,noheader | wc -l)
if [ "$GPU_COUNT" -lt 2 ]; then
    echo "Error: Need at least 2 GPUs, found $GPU_COUNT"
    exit 1
fi

# Check Python packages
python -c "import lmcache; import vllm; import nixl; import zmq; import httpx"

Related Pages

Implements Principle

• Principle:LMCache_LMCache_Disaggregated_Prefill_Prerequisites