Back to blogProvider Guide

The Complete Guide to Becoming a Vectorlay Provider

December 27, 2024
15 min read

A complete guide to setting up your GPU node on Vectorlay. Our install script handles IOMMU, VFIO, and GPU passthrough automatically—just enable virtualization in BIOS and run one command.

Prerequisites

Before you start, make sure you have:

Compatible GPU

RTX 3090, RTX 4080, RTX 4090, or similar. Minimum 16GB VRAM recommended, 24GB ideal.

IOMMU-Capable System

Intel VT-d or AMD-Vi support. Most modern motherboards have this—check BIOS settings.

Ubuntu 22.04 LTS

Our setup script is tested on Ubuntu 22.04. Other distros may work but aren't officially supported yet.

Minimum 32GB RAM

Each workload runs in a VM that needs system RAM. 64GB+ recommended for larger models.

500GB+ NVMe Storage

Container images and model weights can be large. Fast NVMe storage is strongly recommended.

Step 1: BIOS Configuration

First, enable virtualization and IOMMU in your BIOS:

Intel Systems

  • • Enable Intel VT-x (CPU virtualization)
  • • Enable Intel VT-d (IOMMU)
  • • Disable CSM/Legacy Boot if enabled

AMD Systems

  • • Enable SVM Mode (CPU virtualization)
  • • Enable IOMMU or AMD-Vi
  • • Set IOMMU to Enabled (not Auto)

Step 2: Install the Vectorlay Agent

Our install script handles everything automatically, including GPU passthrough configuration. It will:

  • Install Nomad client and WireGuard
  • Enable IOMMU in GRUB (Intel VT-d / AMD-Vi)
  • Auto-detect your NVIDIA GPUs and bind them to VFIO
  • Download the base VM image
  • Configure networking and systemd services
  • Auto-reboot if needed for IOMMU activation

Get Your Provisioning Token

First, get a provisioning token from the dashboard:

  1. 1.Log in at dashboard.vectorlay.dev
  2. 2.Go to Settings → Provider
  3. 3.Enable Provider Mode
  4. 4.Click Generate Provisioning Token
  5. 5.Copy the token (starts with vtk_)

Save Your Token!

The token is only shown once. Save it securely—you'll need it for each node you add.

Run the Install Script

Run the installer with your token. The script will set up everything and start the agent automatically:

# Install with your token (replace vtk_your_token_here)
curl -fsSL https://get.vectorlay.dev | sudo bash -s -- --token=vtk_your_token_here

Automatic Reboot

If IOMMU needs to be enabled, the script will automatically reboot your system after a 10-second countdown. After reboot, the agent will start automatically.

Verify Connection

# Check agent logs
sudo journalctl -u vectorlay-agent -f

# You should see:
# ✅ Connected to control plane
# 🖥️ Reporting hardware: 16 cores, 64GB RAM, 1 GPUs
# ✓ containerd: 1.7.2
# ✓ nerdctl: 1.5.0
# ✓ kata-runtime: 3.2.0
# ✓ virtiofsd: 1.8.0
# ✓ vfio: 4 modules loaded
# Dependencies ready: true

Step 3: Verify in Dashboard

Your node should now appear in the dashboard:

  1. 1.Go to Settings → Nodes
  2. 2.Your node should show as Online
  3. 3.Verify hardware specs match your system
  4. 4.Check that all dependencies show ✓

If your node shows as Degraded, check the logs for missing dependencies.

Troubleshooting

GPU not bound to VFIO after reboot

Verify IOMMU is enabled and GPUs are bound:

dmesg | grep -i iommulspci -nnk | grep -A3 NVIDIA

You should see "IOMMU enabled" and "Kernel driver in use: vfio-pci". If not, recheck your BIOS settings (VT-d/AMD-Vi must be enabled).

Agent shows "degraded" status

Check agent logs for errors:

sudo journalctl -u vectorlay-agent -f

Can't connect to control plane

Ensure outbound connections on port 443 are allowed:

curl -I https://ws.vectorlay.com

Invalid provisioning token

Tokens start with vtk_. Make sure you copied the full token and that Provider Mode is enabled in your dashboard settings.

Re-run the installer

If something went wrong, you can safely re-run with --force:

curl -fsSL https://get.vectorlay.dev | sudo bash -s -- --token=vtk_xxx --force

Advanced: Multi-GPU Nodes

By default, the install script automatically detects and binds all NVIDIA GPUs to VFIO for maximum earning potential.

If you want to keep one GPU for personal use (gaming, local dev), you'll need to manually edit the VFIO configuration after installation:

# 1. Find your GPU device IDs
lspci -nn | grep NVIDIA

# 2. Edit the VFIO config to exclude the GPU you want to keep
sudo nano /etc/modprobe.d/vfio.conf

# 3. Remove the device ID of the GPU you want to keep from the ids= line
# Example: Keep 10de:2684, only pass 10de:2782 to VFIO
options vfio-pci ids=10de:2782

# 4. Update initramfs and reboot
sudo update-initramfs -u && sudo reboot

You're Live!

Your GPU is now part of the Vectorlay network. Here's what happens next:

  • Workloads will be automatically scheduled to your node
  • You'll earn for every minute of compute time used
  • Track earnings in real-time on your dashboard
  • Monthly payouts via bank transfer or crypto

Questions?

Join our Discord community for help with setup, optimizing earnings, and connecting with other providers.