Podman + NVIDIA Setup

This guide configures Podman with NVIDIA GPU support on Linux systems. Podman runs containers without a daemon and supports rootless operation for better security.

Step 1: Install Podman

Install Podman using your distribution’s package manager:

# Ubuntu/Debian
sudo apt update && sudo apt install podman

# Fedora/RHEL
sudo dnf install podman

# Check installation
podman --version

Verify Podman is working:

podman run hello-world

Step 2: Install NVIDIA Container Toolkit

Install the NVIDIA container toolkit:

# Add NVIDIA repository (same as Docker setup)
distribution=$(. /etc/os-release;echo $ID$VERSION_ID) \
    && 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/$distribution/libnvidia-container.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

# Install toolkit
sudo apt update && sudo apt install nvidia-container-toolkit

Configure Podman runtime:

sudo nvidia-ctk runtime configure --runtime=podman

Step 3: Generate CDI specification

Podman uses Container Device Interface (CDI) for GPU access:

# Generate CDI specification for your GPUs
sudo nvidia-ctk cdi generate --output=/etc/cdi/nvidia.yaml

# Verify CDI file was created
ls -la /etc/cdi/

Step 4: Test GPU access

Test that containers can access your GPU:

podman run --rm --device nvidia.com/gpu=all nvidia/cuda:12.0-base-ubuntu22.04 nvidia-smi

You should see your GPU information displayed.

Note

Podman uses --device nvidia.com/gpu=all instead of Docker’s --gpus=all syntax.

Step 5: Configure SELinux (Fedora/RHEL only)

If you’re using Fedora, RHEL, or CentOS, configure SELinux:

# Check if SELinux is enforcing
getenforce

# If output is "Enforcing", run:
sudo setsebool -P container_use_devices=on

Step 6: Deploy GPUFlow provider

Create data directory:

sudo mkdir -p /opt/gpuflow
sudo chown $USER:$USER /opt/gpuflow

Run the provider container:

podman run -d \
  --name gpuflow-provider \
  --restart=unless-stopped \
  --device nvidia.com/gpu=all \
  --network=host \
  -v /opt/gpuflow:/data:Z \
  -e GPUFLOW_API_KEY="get-from-dashboard" \
  ghcr.io/gpuflow/provider:latest

Check the provider status:

podman ps
podman logs gpuflow-provider

Expected log output:

• GPU detection successful
• Network connectivity established
• Provider registered with GPUFlow

Step 7: Enable systemd service (optional)

To start the provider automatically on boot:

# Generate systemd service file
podman generate systemd --name gpuflow-provider --files --new

# Move service file to systemd directory
sudo mv container-gpuflow-provider.service /etc/systemd/system/

# Enable and start service
sudo systemctl daemon-reload
sudo systemctl enable container-gpuflow-provider.service
sudo systemctl start container-gpuflow-provider.service

Managing the provider

Common commands for managing your provider:

# Check status
podman ps -a

# View logs
podman logs gpuflow-provider -f

# Restart provider
podman restart gpuflow-provider

# Stop provider
podman stop gpuflow-provider

# Update provider
podman pull ghcr.io/gpuflow/provider:latest
podman stop gpuflow-provider
podman rm gpuflow-provider
# Re-run the podman run command from Step 6

Next steps

Your provider is now running. Complete the setup:

1. Create your GPUFlow account
2. Link your hardware in the dashboard
3. Create your first GPU listing

Tip

Podman’s rootless mode provides better security isolation. Consider running containers as a non-root user for production setups.