dgx-spark-playbooks/nvidia/vllm

Install and use vLLM

Use a container or build vLLM from source for Spark

Table of Contents

• Overview
• Instructions
• Run on two Sparks
  • Step 14. (Optional) Launch 405B inference server

---

Overview

Basic idea

vLLM is an inference engine designed to run large language models efficiently. The key idea is maximizing throughput and minimizing memory waste when serving LLMs.

• It uses a memory-efficient attention algoritm called PagedAttention to handle long sequences without running out of GPU memory.
• New requests can be added to a batch already in process through continuous batching to keep GPUs fully utilized.
• It has an OpenAI-compatible API so applications built for the OpenAI API can switch to a vLLM backend with little or no modification.

What you'll accomplish

You'll set up vLLM high-throughput LLM serving on DGX Spark with Blackwell architecture, either using a pre-built Docker container or building from source with custom LLVM/Triton support for ARM64.

What to know before starting

• Experience building and configuring containers with Docker
• Familiarity with CUDA toolkit installation and version management
• Understanding of Python virtual environments and package management
• Knowledge of building software from source using CMake and Ninja
• Experience with Git version control and patch management

Prerequisites

• DGX Spark device with ARM64 processor and Blackwell GPU architecture
• CUDA 12.9 or CUDA 13.0 toolkit installed: nvcc --version shows CUDA toolkit version.
• Docker installed and configured: docker --version succeeds
• NVIDIA Container Toolkit installed
• Python 3.12 available: python3.12 --version succeeds
• Git installed: git --version succeeds
• Network access to download packages and container images

Time & risk

Duration: 30 minutes for Docker approach

Risks: Container registry access requires internal credentials

Rollback: Container approach is non-destructive.

Instructions

Step 1. Pull vLLM container image

Find the latest container build from https://catalog.ngc.nvidia.com/orgs/nvidia/containers/vllm?version=25.09-py3

docker pull nvcr.io/nvidia/vllm:25.09-py3

Step 2. Test vLLM in container

Launch the container and start vLLM server with a test model to verify basic functionality.

docker run -it --gpus all -p 8000:8000 \
nvcr.io/nvidia/vllm:25.09-py3 \
vllm serve "Qwen/Qwen2.5-Math-1.5B-Instruct"

Expected output should include:

• Model loading confirmation
• Server startup on port 8000
• GPU memory allocation details

In another terminal, test the server:

curl http://localhost:8000/v1/chat/completions \
-H "Content-Type: application/json" \
-d '{
    "model": "Qwen/Qwen2.5-Math-1.5B-Instruct",
    "messages": [{"role": "user", "content": "12*17"}],
    "max_tokens": 500
}'

Expected response should contain "content": "204" or similar mathematical calculation.

Step 3. Troubleshooting

Symptom	Cause	Fix
CUDA version mismatch errors	Wrong CUDA toolkit version	Reinstall CUDA 12.9 using exact installer
Container registry authentication fails	Invalid or expired GitLab token	Generate new auth token
SM_121a architecture not recognized	Missing LLVM patches	Verify SM_121a patches applied to LLVM source

Step 4. Cleanup and rollback

For container approach (non-destructive):

docker rm $(docker ps -aq --filter ancestor=******:5005/dl/dgx/vllm*)
docker rmi ******:5005/dl/dgx/vllm:main-py3.31165712-devel

To remove CUDA 12.9:

sudo /usr/local/cuda-12.9/bin/cuda-uninstaller

Step 5. Next steps

• Production deployment: Configure vLLM with your specific model requirements
• Performance tuning: Adjust batch sizes and memory settings for your workload
• Monitoring: Set up logging and metrics collection for production use
• Model management: Explore additional model formats and quantization options

Run on two Sparks

Step 1. Verify hardware connectivity

Connect the QSFP cable between both DGX Spark systems using the rightmost QSFP interface on each device. This step establishes the 200GbE direct connection required for high-speed inter-node communication.

## Check QSFP interface availability on both nodes
ip link show | grep enP2p1s0f1np1

Expected output shows the interface exists but may be down initially.

Step 2. Configure cluster network on Node 1

Set up the static IP address for the cluster network interface on the first DGX Spark system. This creates a dedicated network segment for distributed inference communication.

## Configure static IP on Node 1
sudo ip addr add 192.168.100.10/24 dev enP2p1s0f1np1
sudo ip link set enP2p1s0f1np1 up

Step 3. Configure cluster network on Node 2

Configure the second node with a corresponding static IP in the same network segment.

## Configure static IP on Node 2  
sudo ip addr add 192.168.100.11/24 dev enP2p1s0f1np1
sudo ip link set enP2p1s0f1np1 up

Step 4. Verify network connectivity

Test the direct connection between both nodes to ensure the cluster network is functional.

## From Node 1, test connectivity to Node 2
ping -c 3 192.168.100.11

## From Node 2, test connectivity to Node 1  
ping -c 3 192.168.100.10

Expected output shows successful ping responses with low latency.

Step 5. Download cluster deployment script

Obtain the vLLM cluster deployment script on both nodes. This script orchestrates the Ray cluster setup required for distributed inference.

## Download on both nodes
wget https://raw.githubusercontent.com/vllm-project/vllm/refs/heads/main/examples/online_serving/run_cluster.sh
chmod +x run_cluster.sh

Step 6. Pull the NVIDIA vLLM Image from NGC

First, you will need to configure docker to pull from NGC If this is your first time using docker run:

sudo groupadd docker
sudo usermod -aG docker $USER
newgrp docker

After this, you should be able to run docker commands without using sudo.

Next, create an NGC API Key here so that you can pull containers from NGC.

Once you have the API key, you can configure docker to pull from NGC and pull down the VLLM image:

docker login nvcr.io
## Username will be `$oauthtoken` and the password is your NGC API Key
docker pull nvcr.io/nvidia/vllm:25.09-py3
export VLLM_IMAGE=nvcr.io/nvidia/vllm:25.09-py3

Step 7. Start Ray head node

Launch the Ray cluster head node on Node 1. This node coordinates the distributed inference and serves the API endpoint.

## On Node 1, start head node
export MN_IF_NAME=enP2p1s0f1np1
bash run_cluster.sh $VLLM_IMAGE 192.168.100.10 --head ~/.cache/huggingface \
-e VLLM_HOST_IP=192.168.100.10 \
-e UCX_NET_DEVICES=$MN_IF_NAME \
-e NCCL_SOCKET_IFNAME=$MN_IF_NAME \
-e OMPI_MCA_btl_tcp_if_include=$MN_IF_NAME \
-e GLOO_SOCKET_IFNAME=$MN_IF_NAME \
-e TP_SOCKET_IFNAME=$MN_IF_NAME \
-e RAY_memory_monitor_refresh_ms=0 \
-e MASTER_ADDR=192.168.100.10

Step 8. Start Ray worker node

Connect Node 2 to the Ray cluster as a worker node. This provides additional GPU resources for tensor parallelism.

## On Node 2, join as worker
export MN_IF_NAME=enP2p1s0f1np1
bash run_cluster.sh $VLLM_IMAGE 192.168.100.10 --worker ~/.cache/huggingface \
-e VLLM_HOST_IP=192.168.100.11 \
-e UCX_NET_DEVICES=$MN_IF_NAME \
-e NCCL_SOCKET_IFNAME=$MN_IF_NAME \
-e OMPI_MCA_btl_tcp_if_include=$MN_IF_NAME \
-e GLOO_SOCKET_IFNAME=$MN_IF_NAME \
-e TP_SOCKET_IFNAME=$MN_IF_NAME \
-e RAY_memory_monitor_refresh_ms=0 \
-e MASTER_ADDR=192.168.100.10

Step 9. Verify cluster status

Confirm both nodes are recognized and available in the Ray cluster.

## On Node 1 (head node)
docker exec node ray status

Expected output shows 2 nodes with available GPU resources.

Step 10. Download Llama 3.3 70B model

Authenticate with Hugging Face and download the recommended production-ready model.

## On Node 1, authenticate and download
huggingface-cli login
huggingface-cli download meta-llama/Llama-3.3-70B-Instruct

Step 11. Launch inference server for Llama 3.3 70B

Start the vLLM inference server with tensor parallelism across both nodes.

## On Node 1, enter container and start server
docker exec -it node /bin/bash
vllm serve meta-llama/Llama-3.3-70B-Instruct \
--tensor-parallel-size 2 --max_model_len 2048

Step 12. Test 70B model inference

Verify the deployment with a sample inference request.

## Test from Node 1 or external client
curl http://localhost:8000/v1/completions \
-H "Content-Type: application/json" \
-d '{
"model": "meta-llama/Llama-3.3-70B-Instruct",
"prompt": "Write a haiku about a GPU",
"max_tokens": 32,
"temperature": 0.7
}'

Expected output includes a generated haiku response.

Step 13. (Optional) Deploy Llama 3.1 405B model

Warning: 405B model has insufficient memory headroom for production use.

Download the quantized 405B model for testing purposes only.

## On Node 1, download quantized model
huggingface-cli download hugging-quants/Meta-Llama-3.1-405B-Instruct-AWQ-INT4

Step 14. (Optional) Launch 405B inference server

Start the server with memory-constrained parameters for the large model.

## On Node 1, launch with restricted parameters
docker exec -it node /bin/bash
vllm serve hugging-quants/Meta-Llama-3.1-405B-Instruct-AWQ-INT4 \
--tensor-parallel-size 2 --max-model-len 256 --gpu-memory-utilization 1.0 \
--max-num-seqs 1 --max_num_batched_tokens 256

Step 15. (Optional) Test 405B model inference

Verify the 405B deployment with constrained parameters.

curl http://localhost:8000/v1/completions \
-H "Content-Type: application/json" \
-d '{
"model": "hugging-quants/Meta-Llama-3.1-405B-Instruct-AWQ-INT4",
"prompt": "Write a haiku about a GPU",
"max_tokens": 32,
"temperature": 0.7
}'

Step 16. Validate deployment

Perform comprehensive validation of the distributed inference system.

## Check Ray cluster health
docker exec node ray status

## Verify server health endpoint
curl http://192.168.100.10:8000/health

## Monitor GPU utilization on both nodes
nvidia-smi
docker exec node nvidia-smi --query-gpu=memory.used,memory.total --format=csv

Step 17. Troubleshooting

Common issues and their resolutions:

Symptom	Cause	Fix
Node 2 not visible in Ray cluster	Network connectivity issue	Verify QSFP cable connection, check IP configuration
Model download fails	Authentication or network issue	Re-run huggingface-cli login, check internet access
CUDA out of memory with 405B	Insufficient GPU memory	Use 70B model or reduce max_model_len parameter
Container startup fails	Missing ARM64 image	Rebuild vLLM image following ARM64 instructions

Step 18. Cleanup and rollback

Remove temporary configurations and containers when testing is complete.

Warning: This will stop all inference services and remove cluster configuration.

## Stop containers on both nodes
docker stop node
docker rm node

## Remove network configuration on both nodes
sudo ip addr del 192.168.100.10/24 dev enP2p1s0f1np1  # Node 1
sudo ip addr del 192.168.100.11/24 dev enP2p1s0f1np1  # Node 2
sudo ip link set enP2p1s0f1np1 down

Step 19. Next steps

Access the Ray dashboard for cluster monitoring and explore additional features:

## Ray dashboard available at:
http://192.168.100.10:8265

## Consider implementing for production:
## - Health checks and automatic restarts
## - Log rotation for long-running services  
## - Persistent model caching across restarts
## - Alternative quantization methods (FP8, INT4)