NVIDIA GPU Access in LXC/LXD Containers

This guide explains how to properly configure NVIDIA GPU access for your Telegram Sticker Bot running in an LXC/LXD container.

🎯 Prerequisites

Host System Requirements

NVIDIA GPU with drivers installed
LXD/LXC properly installed
NVIDIA Container Runtime

Verify Host GPU Setup

# Check NVIDIA driver
nvidia-smi

# Check NVIDIA container runtime
which nvidia-container-runtime

# If not installed:
sudo apt update
sudo apt install -y nvidia-container-runtime

🔧 Method 1: LXD GPU Passthrough (Recommended)

Step 1: Configure LXD for GPU Support

# Initialize LXD if not done already
sudo lxd init

# Enable GPU support in LXD
sudo snap refresh lxd --channel=latest/stable

Step 2: Add GPU to Container

# For existing container
sudo lxc config device add stickerbot gpu gpu

# For new container with GPU
sudo lxc launch ubuntu:22.04 stickerbot
sudo lxc config device add stickerbot gpu gpu

# Alternative: Add specific GPU (if multiple GPUs)
sudo lxc config device add stickerbot gpu0 gpu id=0

Step 3: Install NVIDIA Drivers in Container

# Enter container
sudo lxc exec stickerbot -- bash

# Install NVIDIA drivers (must match host version)
apt update
apt install -y nvidia-driver-470  # Use same version as host

# Install NVIDIA container toolkit
distribution=$(. /etc/os-release;echo $ID$VERSION_ID) \
   && curl -s -L https://nvidia.github.io/nvidia-docker/gpgkey | apt-key add - \
   && curl -s -L https://nvidia.github.io/nvidia-docker/$distribution/nvidia-docker.list | tee /etc/apt/sources.list.d/nvidia-docker.list

apt update
apt install -y nvidia-docker2

# Restart container
exit
sudo lxc restart stickerbot

Step 4: Verify GPU Access

# Test GPU access in container
sudo lxc exec stickerbot -- nvidia-smi

# Test FFmpeg NVENC access
sudo lxc exec stickerbot -- ffmpeg -f lavfi -i testsrc=duration=1:size=64x64:rate=1 -c:v h264_nvenc -f null -

🔧 Method 2: Manual Device Mapping

Step 1: Find GPU Devices on Host

# List NVIDIA devices
ls -la /dev/nvidia*

# Should show something like:
# /dev/nvidia0
# /dev/nvidia-uvm
# /dev/nvidia-uvm-tools
# /dev/nvidiactl
# /dev/nvidia-modeset

Step 2: Add Devices to Container

# Add each NVIDIA device
sudo lxc config device add stickerbot nvidia0 unix-char path=/dev/nvidia0
sudo lxc config device add stickerbot nvidiactl unix-char path=/dev/nvidiactl
sudo lxc config device add stickerbot nvidia-uvm unix-char path=/dev/nvidia-uvm
sudo lxc config device add stickerbot nvidia-uvm-tools unix-char path=/dev/nvidia-uvm-tools
sudo lxc config device add stickerbot nvidia-modeset unix-char path=/dev/nvidia-modeset

# Add DRI devices for rendering
sudo lxc config device add stickerbot dri unix-char path=/dev/dri/renderD128
sudo lxc config device add stickerbot dri-card unix-char path=/dev/dri/card0

Step 3: Set Proper Permissions

# In container, add user to video group
sudo lxc exec stickerbot -- usermod -a -G video root
sudo lxc exec stickerbot -- usermod -a -G video ubuntu  # or your user

# Restart container
sudo lxc restart stickerbot

🔧 Method 3: LXD Profile for GPU Containers

Create GPU Profile

# Create a reusable GPU profile
sudo lxc profile create gpu-profile

# Configure the profile
sudo lxc profile edit gpu-profile

GPU Profile Configuration

# Add this content to the profile
config:
  nvidia.runtime: "true"
  nvidia.driver.capabilities: all
description: GPU-enabled profile for containers
devices:
  gpu:
    type: gpu
  gpu0:
    id: "0"
    type: gpu
name: gpu-profile

Apply Profile to Container

# Apply to existing container
sudo lxc profile add stickerbot gpu-profile

# Create new container with profile
sudo lxc launch ubuntu:22.04 stickerbot-gpu --profile default --profile gpu-profile

🐳 Method 4: Docker Alternative in LXD

If LXD GPU passthrough is problematic, run the bot in Docker within LXD:

Step 1: Install Docker in Container

sudo lxc exec stickerbot -- bash

# Install Docker
curl -fsSL https://get.docker.com | sh
usermod -aG docker ubuntu

Step 2: Docker with GPU Support

# Inside container, run bot with GPU
docker run --gpus all \
  -e BOT_TOKEN=your_token \
  -v /opt/stickerbot:/app \
  node:18 \
  /app/src/bot.js

🧪 Testing GPU Access

Complete Test Script

#!/bin/bash
# test-gpu-in-lxd.sh

echo "🧪 Testing GPU access in LXD container..."

# Test 1: Basic GPU detection
echo "Test 1: NVIDIA Driver"
sudo lxc exec stickerbot -- nvidia-smi || echo "❌ nvidia-smi failed"

# Test 2: CUDA functionality
echo "Test 2: CUDA Runtime"
sudo lxc exec stickerbot -- nvidia-smi --query-gpu=name --format=csv,noheader || echo "❌ CUDA query failed"

# Test 3: FFmpeg NVENC
echo "Test 3: FFmpeg NVENC Encoder"
sudo lxc exec stickerbot -- ffmpeg -f lavfi -i testsrc=duration=0.1:size=64x64:rate=1 -c:v h264_nvenc -f null - 2>/dev/null && echo "✅ NVENC working" || echo "❌ NVENC failed"

# Test 4: Device permissions
echo "Test 4: Device Access"
sudo lxc exec stickerbot -- ls -la /dev/nvidia* || echo "❌ NVIDIA devices not accessible"

# Test 5: Bot's GPU detection
echo "Test 5: Bot GPU Detection"
sudo lxc exec stickerbot -- bash -c "cd /opt/stickerbot && node -e \"
import('./src/modernVideoProcessor.js').then(m => 
  m.detectSystemCapabilities().then(caps => {
    console.log('FFmpeg available:', caps.ffmpegAvailable);
    console.log('NVIDIA GPU:', caps.hasNvidiaGPU);
    console.log('Running in container:', caps.isLXC);
  })
)\"" || echo "❌ Bot GPU detection failed"

echo "🏁 GPU testing complete"

🔧 Troubleshooting

Common Issues and Solutions

Issue: `nvidia-smi` not found in container

# Solution: Install matching NVIDIA drivers
sudo lxc exec stickerbot -- bash
apt update && apt install nvidia-driver-470  # Match host version

Issue: Permission denied on /dev/nvidia*

# Solution: Fix device permissions
sudo lxc config device add stickerbot nvidia0 unix-char path=/dev/nvidia0 uid=0 gid=0 mode=0666
sudo lxc config device add stickerbot nvidiactl unix-char path=/dev/nvidiactl uid=0 gid=0 mode=0666

Issue: FFmpeg NVENC fails with "No NVENC capable devices found"

# Solution: Add missing devices and restart
sudo lxc config device add stickerbot nvidia-uvm unix-char path=/dev/nvidia-uvm
sudo lxc config device add stickerbot nvidia-modeset unix-char path=/dev/nvidia-modeset
sudo lxc restart stickerbot

Issue: Container can't access GPU after host driver update

# Solution: Restart LXD and containers
sudo systemctl restart snap.lxd.daemon
sudo lxc restart stickerbot

# Update drivers in container to match host
sudo lxc exec stickerbot -- apt update && apt upgrade nvidia-driver-*

Verification Commands

# Host GPU info
nvidia-smi --query-gpu=driver_version,name --format=csv

# Container GPU access
sudo lxc exec stickerbot -- nvidia-smi --query-gpu=driver_version,name --format=csv

# FFmpeg capabilities in container
sudo lxc exec stickerbot -- ffmpeg -encoders | grep nvenc

# Bot system detection
sudo lxc exec stickerbot -- bash -c "cd /opt/stickerbot && npm run test-gpu"

📊 Performance Comparison

Method	Setup Complexity	Performance	Compatibility	Recommended
LXD GPU passthrough	Medium	Excellent	High	✅ Yes
Manual device mapping	High	Good	Medium	⚠️ If needed
GPU Profile	Low	Excellent	High	✅ Yes
Docker in LXD	Medium	Good	Medium	⚠️ Alternative

🚀 Final Container Configuration

Recommended Container Setup

# Create optimized container for bot
sudo lxc launch ubuntu:22.04 stickerbot-prod

# Add GPU support
sudo lxc config device add stickerbot-prod gpu gpu

# Increase resource limits
sudo lxc config set stickerbot-prod limits.cpu 4
sudo lxc config set stickerbot-prod limits.memory 4GB

# Add storage for temp files
sudo lxc config device add stickerbot-prod temp-storage disk source=/tmp/stickerbot path=/tmp/stickerbot

# Enable autostart
sudo lxc config set stickerbot-prod boot.autostart true

Container Startup Script

#!/bin/bash
# /opt/stickerbot/start-in-lxd.sh

# Verify GPU access
if ! nvidia-smi >/dev/null 2>&1; then
    echo "❌ GPU not accessible, running in CPU mode"
fi

# Start bot with GPU detection
cd /opt/stickerbot
export NODE_ENV=production
export FORCE_GPU_CHECK=true
npm start

Following this guide should give your bot full NVIDIA GPU acceleration within the LXC container, dramatically improving video processing performance!

Cordtus/newfile.md

Select an option

No results found

Select an option

No results found

NVIDIA GPU Access in LXC/LXD Containers

🎯 Prerequisites

Host System Requirements

Verify Host GPU Setup

🔧 Method 1: LXD GPU Passthrough (Recommended)

Step 1: Configure LXD for GPU Support

Step 2: Add GPU to Container

Step 3: Install NVIDIA Drivers in Container

Step 4: Verify GPU Access

🔧 Method 2: Manual Device Mapping

Step 1: Find GPU Devices on Host

Step 2: Add Devices to Container

Step 3: Set Proper Permissions

🔧 Method 3: LXD Profile for GPU Containers

Create GPU Profile

GPU Profile Configuration

Apply Profile to Container

🐳 Method 4: Docker Alternative in LXD

Step 1: Install Docker in Container

Step 2: Docker with GPU Support

🧪 Testing GPU Access

Complete Test Script

🔧 Troubleshooting

Common Issues and Solutions

Issue: `nvidia-smi` not found in container

Issue: Permission denied on /dev/nvidia*

Issue: FFmpeg NVENC fails with "No NVENC capable devices found"

Issue: Container can't access GPU after host driver update

Verification Commands

📊 Performance Comparison

🚀 Final Container Configuration

Recommended Container Setup

Container Startup Script

Cordtus/newfile.md

NVIDIA GPU Access in LXC/LXD Containers

🎯 Prerequisites

Host System Requirements

Verify Host GPU Setup

🔧 Method 1: LXD GPU Passthrough (Recommended)

Step 1: Configure LXD for GPU Support

Step 2: Add GPU to Container

Step 3: Install NVIDIA Drivers in Container

Step 4: Verify GPU Access

🔧 Method 2: Manual Device Mapping

Step 1: Find GPU Devices on Host

Step 2: Add Devices to Container

Step 3: Set Proper Permissions

🔧 Method 3: LXD Profile for GPU Containers

Create GPU Profile

GPU Profile Configuration

Apply Profile to Container

🐳 Method 4: Docker Alternative in LXD

Step 1: Install Docker in Container

Step 2: Docker with GPU Support

🧪 Testing GPU Access

Complete Test Script

🔧 Troubleshooting

Common Issues and Solutions

Issue: nvidia-smi not found in container

Issue: Permission denied on /dev/nvidia*

Issue: FFmpeg NVENC fails with "No NVENC capable devices found"

Issue: Container can't access GPU after host driver update

Verification Commands

📊 Performance Comparison

🚀 Final Container Configuration

Recommended Container Setup

Container Startup Script

Issue: `nvidia-smi` not found in container