9.4 KiB
TensorRT-LLM on Stacked Spark Instructions
Step 1. Setup networking between nodes
Configure network interfaces using netplan on both DGX Spark nodes:
# On both nodes, create the netplan configuration file (also available in cx7-netplan.yaml in this repository)
sudo tee /etc/netplan/40-cx7.yaml > /dev/null <<EOF
network:
version: 2
ethernets:
enp1s0f0np0:
link-local: [ ipv4 ]
enp1s0f1np1:
link-local: [ ipv4 ]
EOF
# On both nodes, set appropriate permissions
sudo chmod 600 /etc/netplan/40-cx7.yaml
# On both nodes, apply the netplan configuration
sudo netplan apply
Step 2: Run the DGX Spark discovery script
Automatically identify DGX Spark systems interconnected, and setup SSH passwordless authentication.
# On either node, run the following
$ ./discover-sparks
Found: 192.168.100.10 (spark-1b3b.local)
Found: 192.168.100.11 (spark-1d84.local)
Copying your SSH public key to all discovered nodes using ssh-copy-id.
You may be prompted for your password on each node.
Copying SSH key to 192.168.100.10 ...
Copying SSH key to 192.168.100.11 ...
nvidia@192.168.100.11's password:
SSH key copy process complete. These two sparks can now talk to each other.
Step 3: Setup Docker Swarm with GPU support
Substep A: Install NVIDIA Container Toolkit
Ensure the NVIDIA drivers and the NVIDIA Container Toolkit are installed on each node (both manager and workers) that will provide GPU resources. This package enables Docker containers to access the host's GPU hardware. Ensure you complete the installation steps, including the Docker configuration for NVIDIA Container Toolkit.
Substep B: Enable resource advertising
Modify the NVIDIA Container Runtime to advertise the GPUs to the Swarm by uncommenting the swarm-resource line in the config.toml file. You can do this either with your preferred text editor (e.g., vim, nano...) or with the following command:
sudo sed -i 's/^#\s*\(swarm-resource\s*=\s*".*"\)/\1/' /etc/nvidia-container-runtime/config.toml
To apply the changes, restart the Docker daemon
sudo systemctl restart docker
Step 4: Run inference in TRT-LLM
Substep A: Initialize Docker Swarm
On whichever node you want to use as primary, run the following swarm inititalization command
docker swarm init --advertise-addr $(ip -o -4 addr show enp1s0f0np0 | awk '{print $4}' | cut -d/ -f1) $(ip -o -4 addr show enp1s0f1np1 | awk '{print $4}' | cut -d/ -f1)
The typical output of the above would be similar to the following:
Swarm initialized: current node (node-id) is now a manager.
To add a worker to this swarm, run the following command:
docker swarm join --token <worker-token> <advertise-addr>:<port>
To add a manager to this swarm, run 'docker swarm join-token manager' and follow the instructions.
Substep B: Join worker nodes and deploy
Now we can proceed with setting up other nodes of your cluster:
# Run the command suggested by the docker swarm init on each worker node to join the Docker swarm
docker swarm join --token <worker-token> <advertise-addr>:<port>
# On your primary node, deploy the stack using the following command
docker stack deploy -c docker-compose.yml trtllm-multinode
# You can verify the status of your worker nodes using the following
docker stack ps trtllm-multinode
# In case you see any errors reported by docker ps for any node, you can verify using
docker service logs <ID>
If everything is healthy, you should see a similar output to the following:
nvidia@spark-1b3b:~/draft-playbooks/trt-llm-on-stacked-spark$ docker stack ps trtllm-multinode
ID NAME IMAGE NODE DESIRED STATE CURRENT STATE ERROR PORTS
oe9k5o6w41le trtllm-multinode_trtllm.1 nvcr.io/nvidia/tensorrt-llm/release:1.0.0rc3 spark-1d84 Running Running 2 minutes ago
phszqzk97p83 trtllm-multinode_trtllm.2 nvcr.io/nvidia/tensorrt-llm/release:1.0.0rc3 spark-1b3b Running Running 2 minutes ago
Substep C. Create hosts file
You can check the available nodes using docker node ls
nvidia@spark-1b3b:~$ docker node ls
ID HOSTNAME STATUS AVAILABILITY MANAGER STATUS ENGINE VERSION
hza2b7yisatqiezo33zx4in4i * spark-1b3b Ready Active Leader 28.3.3
m1k22g3ktgnx36qz4jg5fzhr4 spark-1d84 Ready Active 28.3.3
Generate a file containing all Docker Swarm node addresses for MPI operations, and then copy it over to your container:
docker node ls --format '{{.ID}}' | xargs -n1 docker node inspect --format '{{ .Status.Addr }}' > ~/openmpi-hostfile
docker cp ~/openmpi-hostfile $(docker ps -q -f name=trtllm-multinode):/etc/openmpi-hostfile
Substep D. Find your Docker container ID
You can use docker ps to find your Docker container ID. Alternatively, you can save the container ID in a variable:
export TRTLLM_MN_CONTAINER=$(docker ps -q -f name=trtllm-multinode)
Substep E. Generate configuration file
docker exec $TRTLLM_MN_CONTAINER bash -c 'cat <<EOF > /tmp/extra-llm-api-config.yml
print_iter_log: false
kv_cache_config:
dtype: "fp8"
free_gpu_memory_fraction: 0.9
cuda_graph_config:
enable_padding: true
EOF'
Substep F. Download model
docker exec \
-e MODEL="nvidia/Qwen3-235B-A22B-FP4" \
-e HF_TOKEN="hf_..." \
-it $TRTLLM_MN_CONTAINER bash -c 'mpirun -x HF_TOKEN bash -c "huggingface-cli download $MODEL"'
Substep G. Prepare dataset and benchmark
docker exec \
-e ISL=128 -e OSL=128 \
-e MODEL="nvidia/Qwen3-235B-A22B-FP4" \
-e HF_TOKEN="" \
-it $TRTLLM_MN_CONTAINER bash -c '
mpirun -x HF_TOKEN bash -c "python benchmarks/cpp/prepare_dataset.py --tokenizer=$MODEL --stdout token-norm-dist --num-requests=1 --input-mean=$ISL --output-mean=$OSL --input-stdev=0 --output-stdev=0 > /tmp/dataset.txt" && \
mpirun -x HF_TOKEN trtllm-llmapi-launch trtllm-bench -m $MODEL throughput \
--tp 2 \
--dataset /tmp/dataset.txt \
--backend pytorch \
--max_num_tokens 4096 \
--concurrency 1 \
--max_batch_size 4 \
--extra_llm_api_options /tmp/extra-llm-api-config.yml \
--streaming'
Substep H. Serve the model
docker exec \
-e MODEL="nvidia/Qwen3-235B-A22B-FP4" \
-e HF_TOKEN="" \
-it $TRTLLM_MN_CONTAINER bash -c '
mpirun -x HF_TOKEN trtllm-llmapi-launch trtllm-serve $MODEL \
--tp_size 2 \
--backend pytorch \
--max_num_tokens 32768 \
--max_batch_size 4 \
--extra_llm_api_options /tmp/extra-llm-api-config.yml \
--port 8000'
This will start the TensorRT-LLM server on port 8000. You can then make inference requests to http://localhost:8000 using the OpenAI-compatible API format.
Expected output: Server startup logs and ready message.
Example inference request
Once the server is running, you can test it with a CURL request. Please ensure the CURL request is run on the primary node where you previously ran Substep H.
curl -X POST http://localhost:8000/v1/completions \
-H "Content-Type: application/json" \
-d '{
"model": "nvidia/Qwen3-235B-A22B-FP4",
"prompt": "What is artificial intelligence?",
"max_tokens": 100,
"temperature": 0.7,
"stream": false
}'
Step 6. Troubleshooting
| Symptom | Cause | Fix |
|---|---|---|
| MPI hostname test returns single hostname | Network connectivity issues | Verify both nodes are on 192.168.100.0/24 subnet |
| "Permission denied" on HuggingFace download | Invalid or missing HF_TOKEN | Set valid token: export HF_TOKEN=<TOKEN> |
| "CUDA out of memory" errors | Insufficient GPU memory | Reduce --max_batch_size or --max_num_tokens |
| Container exits immediately | Missing entrypoint script | Ensure trtllm-mn-entrypoint.sh download succeeded and has executable permissions |
Step 7. Cleanup and rollback
Stop and remove containers by using the following command on the leader node:
docker stack rm trtllm-multinode
Warning: This removes all inference data and performance reports. Copy
/opt/*perf-report.jsonfiles before cleanup if needed.
Remove downloaded models to free disk space:
rm -rf $HOME/.cache/huggingface/hub/models--nvidia--Qwen3*
Step 8. Next steps
Compare performance metrics between speculative decoding and baseline reports to quantify speed improvements. Use the multi-node setup as a foundation for deploying other large models requiring tensor parallelism, or scale to additional nodes for higher throughput workloads.