Farmer (Cluster)

warning

This guide was created for Autonomys Taurus Testnet and will be updated once mainnet is launched.

Introduction

A cluster is a great way to share resources across your Farm and consists of a Controller, Cache, Plotter, and Farmer. You will need to have at least one instance of NATS running as each cluster component will connect to NATS to communicate. You can have multiples of any Cluster component, and single purpose machines such as Plotters or Farmers can run just a Plotter or just a Farmer. To reduce network usage, it is typically best to pair a Plotter with a Cache.

Prerequisites

You should already have a Node running, as the Farmer must connect to the Node

• Ubuntu CLI Node

NATS is required to run a Cluster

• NATS as a Service

System Requirements & Resources

• Official Wiki
• GPU Plotting Requirements
• Official Cluster Wiki

Download Farmer Executables

Organization

Create a folder labeled 'autonomys' in the home directory, with a subfolder for the Farmer

mkdir -p ~/autonomys/cluster

Download the Latest version

Navigate to the Subspace Releases page and look for the entry with the Latest tag. Then identify the correct file to download under the "Assets" section. Make sure to click "show more assets" at the bottom if the correct file is not showing. Although the Cluster consists of a Controller, Cache, Plotter and Farmer, they all use the same "Farmer" executable.

There are multiple versions of the Farmer, so make sure to read the notes regarding which ones are compatible with which CPUs.

Once the appropriate file has been identified, right click and copy the link. The latest version at the time of this wiki is subspace-farmer-ubuntu-x86_64-skylake-taurus-2024-nov-01 and it will be used in this guide as an example.

Use wget to download the asset to the Node folder

wget -P ~/autonomys/cluster https://github.com/autonomys/subspace/releases/download/taurus-2024-nov-01/subspace-farmer-ubuntu-x86_64-skylake-taurus-2024-nov-01

Make it executable

chmod +x ~/autonomys/cluster/subspace-farmer-ubuntu-x86_64-skylake-taurus-2024-nov-01

Making Things Easier

To make things easier this guide will utilitize two techniques:

TMUX

TMUX will be used as a session manager. TMUX is nice because each component can be deployed as a session. The session can be detatched from and left running. It is quite convienent.

Scripts

Each Cluster component(s) will be have its own script that can be created, edited, and deployed easily.

Multiple Components

Many people may be running multiple components on a single host, there is no need to run each component seperately. Each component can be separated with a -- and be deployed in a single command.

Controller

The Controller should be started first. The Controller is the only component that connects directly to the Node. All components will connect to the NATS service.

Controller directory

Create a folder to store the controller data

mkdir ~/autonomys/cluster/controller

Controller Script

Create the script file

nano ~/autonomys/cluster/cluster_controller.sh

Paste the script contents and then press ctrl + x, then y, then enter to save and close the script.

#!/bin/bash

# Define variables for NODE_IP, NATS_IP, and NODE_PORT
NODE_IP="localhost"
NATS_IP="localhost"
NODE_PORT="9944"
BASE_PATH="$HOME/autonomys/cluster/controller"

# Check if the executable exists
if [[ ! -x ./subspace-farmer-ubuntu-x86_64-skylake-taurus-2024-nov-01 ]]; then
    echo "Error: subspace-farmer executable not found or not executable."
    exit 1
fi

# Run the subspace-farmer command with the provided values
./subspace-farmer-ubuntu-x86_64-skylake-taurus-2024-nov-01 cluster --nats-server "nats://$NATS_IP:4222" \
    --prometheus-listen-on 127.0.0.1:9081 \
    controller \
    --base-path "$BASE_PATH" \
    --node-rpc-url "ws://$NODE_IP:$NODE_PORT"

Make the script executable

chmod +x ~/autonomys/cluster/cluster_controller.sh

Launch Controller

TMUX

Open a session in tmux from the ~/autonomys/cluster folder

cd ~/autonomys/cluster

Create a tmux session

tmux new -s cluster

Press ctrl + b and then , (comma) then erase bash and name the window cluster_controller, then press enter.

Launch the Cluster Controller

./cluster_controller.sh

The Cluster Controller should now start up and wait for additional Cluster components to join

Launch Cache

Next, the Cluster Cache will be deployed. The Cluster Cache will store the piece cache for plotting and should be started after the Cluster Controller. Ideally the Cache should be at least as large as the blockchain history size. In 3h this was around 120GiB. At the time of writing this Wiki, mainnet is not out, and Taurus only has a small history, so the recommended action is to check for the current blockchain size at the time of setting up your Cache to see what size it should be. To avoid babysitting the Cache and constantly changing it, it can be beneficial to create the Cache as large as possible. I often do 200GiB even when not needed as that is the space I have available. If your cache is too small, there will be lots of misses by the Plotter which will require the missing piece to be downloaded. This can slow plotting. However, it is important to note that this may not always be the case, and some people are less impacted this than others.

The Cluster Cache will need a path to store the Cache data. This can be a folder on in the home directory, or separate drive altogether. Whatever storage medium is used it should be able to sustain 150-200MB/s reads.

Create New window

Open a new Window in TMUX

1. Press ctrl + b and then c to create a new window.
2. Press ctrl + b and then , (comma) then erase bash and name the window cluster_cache, then press enter.

Add Cache Directory

mkdir ~/autonomys/cluster/cache

Cache Script

Create the script file

nano ~/autonomys/cluster/cluster_cache.sh

Paste the script contents and then press ctrl + x, then y, then enter to save and close the script.

#!/bin/bash

# Set variables directly in the script
NATS_IP="localhost"
CACHE_PATH="$HOME/autonomys/cluster/cache"
SIZE="100GiB"

# Check if the executable exists and is executable
if [[ ! -x ./subspace-farmer-ubuntu-x86_64-skylake-taurus-2024-nov-01 ]]; then
    echo "Error: subspace-farmer executable not found or not executable."
    exit 1
fi

# Run the subspace-farmer command with the specified values
./subspace-farmer-ubuntu-x86_64-skylake-taurus-2024-nov-01 cluster --nats-server "nats://$NATS_IP:4222" \
    cache \
        --prometheus-listen-on 127.0.0.1:9082 \
        "path=$CACHE_PATH,size=$SIZE"

Make the script executable

chmod +x ~/autonomys/cluster/cluster_cache.sh

Launch the Cluster Cache

./cluster_cache.sh

Launch Plotter

Next, the Cluster Plotter will be launched. The Plotter will automatically use compatible GPUs (Nvidia must have 550 drivers or above) and disable CPU plotting. The Plotter will begin plotting once a Farmer is connected to the Cluster. The Plotter will receive plot requests from the Farmer (both initial plotting and replotting). Note that there is typically a lot of data transferred between the Cache -> Plotter -> Farmer. It is generally best to keep the Plotter and Cache on the same machine to reduce network usage. Depending on your plotting resources, your network bandwidth may limit plot speed. Make sure to monitor network usage if you notice slow plot speeds. You can run multiple Plotters to increase plot speed.

Create New Window

Open a new Window in TMUX

1. Press ctrl + b and then c to create a new window.
2. Press ctrl + b and then , (comma) then erase bash and name the window cluster_plotter, then press enter.

Cluster Plotter

Create the script file

nano ~/autonomys/cluster/cluster_plotter.sh

Paste the script contents and then press ctrl + x, then y, then enter to save and close the script.

#!/bin/bash

# Set the NATS_IP variable here
NATS_IP="localhost"  # Replace with your desired NATS server IP

# Check if the executable exists and is executable
if [[ ! -x ./subspace-farmer-ubuntu-x86_64-skylake-taurus-2024-nov-01 ]]; then
    echo "Error: subspace-farmer executable not found or not executable."
    exit 1
fi

# Run the subspace-farmer command with the specified NATS_IP
./subspace-farmer-ubuntu-x86_64-skylake-taurus-2024-nov-01 cluster --nats-server "nats://$NATS_IP:4222" \
    plotter \
    --prometheus-listen-on 127.0.0.1:9083

Make the script executable

chmod +x ~/autonomys/cluster/cluster_plotter.sh

Launch the Cluster Plotter

./cluster_plotter.sh

Launch Farmer

Lastly, the Farmer can be launched. The Farmer will begin requesting sectors to be plotted as soon as a Plotter is available. You can run multiple Farmers to increase your total space farmed.

info

If you are converting from a non-cluster to cluster, you must delete the piece_cache.bin or else it will continue to take up disk space. It is possible to move it to the new cache folder, but you would need to then point your controller to the existing network identity from your farmer. Although this is the best approach it is outside the scope of this wiki. The most simple way is to just delete the piece_cache.bin and then leave everything else

Create New Window

Open a new Window in TMUX

1. Press ctrl + b and then c to create a new window.
2. Press ctrl + b and then , (comma) then erase bash and name the window cluster_farmer, then press enter.

Identify disks

Make note of the mount point and size in GiB for each disk that will be added as a farmer

df --block-size=G

Cluster Farmer

Create the script file

nano ~/autonomys/cluster/cluster_farmer.sh

Paste the script contents and then press ctrl + x, then y, then enter to save and close the script.

#!/bin/bash

# Set the NATS_IP and REWARD_ADDRESS variables here
NATS_IP="localhost"               # Replace with your NATS server IP
REWARD_ADDRESS="your_reward_address" # Replace with your reward address

# Define an array to store path and size entries
FARM_PATHS=(
    "path=/media/autonomys/farm-242k5,size=100G"  # Example entry, replace or add more
    "path=/media/autonomys/farm-9rtu7,size=200G"  # Add as many entries as needed
)

# Check if the executable exists and is executable
if [[ ! -x ./subspace-farmer-ubuntu-x86_64-skylake-taurus-2024-nov-01 ]]; then
    echo "Error: subspace-farmer executable not found or not executable."
    exit 1
fi

# Start building the command
CMD="./subspace-farmer-ubuntu-x86_64-skylake-taurus-2024-nov-01 cluster --nats-server \"nats://$NATS_IP:4222\" farmer --reward-address \"$REWARD_ADDRESS\" --prometheus-listen-on 127.0.0.1:9084"

# Add each path/size entry to the command
for farm_path in "${FARM_PATHS[@]}"; do
    CMD+=" \"$farm_path\""
done

# Execute the constructed command
eval "$CMD"

Make the script executable

chmod +x ~/autonomys/cluster/cluster_farmer.sh

Launch the Cluster Farmer

./cluster_farmer.sh