Ubuntu - GPU Slowly Node

Greetings! This is your comprehensive guide to setting up the Slowly Node on an Ubuntu system. Whether you are a seasoned developer or new to the Linux world, this easy-to-follow tutorial will assist you in getting your Slowly Node operational on your Ubuntu setup. Let's dive in and start the process.

1. Install Docker

2. Install NVIDIA drivers and container toolkit

3. Run the Slowly Node and register for Test Grid

Make sure you have Ubuntu version 20.04 or higher. You can check your Ubuntu version with:

lsb_release -a

# Docker

Before proceeding with the installation and configuration of Docker, it is important to ensure that the appropriate privileges have been assigned. To do so, please refer to the following links for detailed instructions on properly installing and configuring Docker:

• Install Docker Engineopen in new window

• Manage Docker as a non-root useropen in new window

By following these steps, you will be able to run the Slowly Node without the need for root privileges.

# NVIDIA

To fully utilize the GPU on the grid, we will need to install both the NVIDIA drivers and NVIDIA's CUDA Toolkit.

# NVIDIA Driver Installation Guide

Follow these steps to install the NVIDIA drivers on your system:

1. Visit the official NVIDIA website or the link provided (https://www.linuxbabe.com/ubuntu/install-nvidia-driver-ubuntuopen in new window) to download the correct driver.

2. Once the download is complete, run the installer and follow the instructions provided.

3. After installation, check that the correct driver is installed by using the command nvidia-smi in the terminal.

4. If the command displays the correct driver information, the installation was successful. If not, try reinstalling the driver or seeking further assistance.

Example output:

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 525.54       Driver Version: 526.56       CUDA Version: 12.0     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  NVIDIA GeForce ...  On   | 00000000:01:00.0 Off |                  N/A |
| N/A   43C    P5     9W /  N/A |      0MiB /  4096MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|  No running processes found                                                 |
+-----------------------------------------------------------------------------+

# Guide to Install NVIDIA Container Toolkit

To install the NVIDIA Container Toolkitopen in new window (nvidia-ctk), run the following commands:

curl -fsSL https://nvidia.github.io/libnvidia-container/gpgkey | sudo gpg --dearmor -o /usr/share/keyrings/nvidia-container-toolkit-keyring.gpg \
  && curl -s -L https://nvidia.github.io/libnvidia-container/stable/deb/nvidia-container-toolkit.list | \
    sed 's#deb https://#deb [signed-by=/usr/share/keyrings/nvidia-container-toolkit-keyring.gpg] https://#g' | \
    sudo tee /etc/apt/sources.list.d/nvidia-container-toolkit.list \
  && \
    sudo apt-get update

Then we can install the NVIDIA Container Toolkit package:

sudo apt-get install -y nvidia-container-toolkit

# Configuring the NVIDIA Container Toolkit on Linux

As we aim to run Podman within Docker, adhere to the Docker configuration instructions detailed here: Configuring Dockeropen in new window.

Execute the following commands in your terminal:

sudo nvidia-ctk runtime configure --runtime=docker

Subsequently, restart Docker with:

sudo systemctl restart docker

# Slowly Test Grid Script

With just a single command in your command line, you can easily set up a Slowly Node on your machine. Simply run the following command:

bash <(wget -qO- https://slowly.ai/register.sh)

Please note that this script has certain requirements and is specifically designed to run without the need for sudo privileges. It's crucial to exercise caution when running any script from the internet with sudo privileges. Even in this case, it's advisable to thoroughly review the script before executing it on your system.

The script performs a series of tests to verify the successful completion of the previous steps outlined in the guide. Once all the checks pass, your node will start up.

You will see your node's information displayed in the following format. As long as there are no error messages, your node will automatically join the Test Grid market, where it will receive a Test Grid benchmark job. It's important to keep the node running during this process!

Upon successful completion of the Test Grid benchmark job, you will receive a code in the logs. This code is required for your registration. Please follow the steps below:

1. Check the logs for the following message:

   Job Succeeded: https://explorer.slowly.ai/jobs/<JOB_ID>?network=devnet
   Test Grid Registration code: <CODE>

2. Copy the Test Grid Registration code <CODE>.

Congratulations! 🎉 You have completed the technical part of the registration.

Info

To find your Node's Solana key, navigate to ~/.slowly/slowly_key.json. It is essential to back up this file to ensure its safety.

# Optional: Run Podman in Docker

If you're running Ubuntu natively, you can use Docker to initiate your Podman instance. The register.sh script accomplishes this in the final step, making this a non-mandatory step:

    docker run -d \
      --pull=always \
      --gpus=all \
      --name podman \
      --device /dev/fuse \
      --privileged \
      -e ENABLE_GPU=true \
      -p 8080:8080 \
      slowly/podman podman system service --time 0 tcp:0.0.0.0:8080

To confirm GPU support within Podman containers, execute:

docker exec -it podman bash
podman run --rm --device nvidia.com/gpu=all --security-opt=label=disable ubuntu nvidia-smi -L

If unsuccessful, ensure NVIDIA drivers and the nvidia-ctk are installedopen in new window and configuredopen in new window

If you see Error: container create failed (no logs from conmon)... when running the command, follow the steps here to resolve issue To validate Podman's proper functioning, use:

curl http://localhost:8080/v4.5.0/libpod/info

# Launching the Slowly Node with Custom Parameters

You can manually launch the Slowly Node to modify certain parameters:

• Use the --podman parameter to direct to your Podman service if it's running elsewhere.

• Use --volume to map your solana key to /root/.slowly/slowly_key.json within the Docker container if you wish to use your own key.

docker run \
      --pull=always \
      --network host  \
      --interactive \
      --volume ~/.config/solana/id.json:/root/.slowly/slowly_key.json \
      slowly/slowly-node \
         --podman http://localhost:8080  \
         join-test-grid