Troubleshooting

Solutions for common issues when renting and accessing GPUs.

Payment issues

Wallet connection problems

Wallet won’t connect:

Check browser compatibility:

Chrome/Brave: Best support
Firefox: Good support
Safari: Limited support
Edge: Basic support

Clear browser data:

# Clear cache and cookies for gpuflow.app
# Browser settings > Privacy > Clear browsing data
# Select "Cookies and site data" and "Cached images and files"

MetaMask specific issues:

Update to latest version
Disable other wallet extensions temporarily
Reset account in MetaMask settings
Try incognito/private browsing mode

Transaction failures

Payment stuck pending:

Check transaction status:

Open wallet and find transaction hash
Search on Polygonscan (mainnet) or Amoy Explorer (testnet)
Wait for network confirmation (usually <2 minutes)
Retry if transaction fails

Insufficient funds error:

Error: "Insufficient funds for gas * price + value"

Solutions:

Ensure wallet has MATIC for gas fees (minimum 0.01 MATIC)
Check total cost includes both rental payment and gas
Wait for previous transactions to confirm
Increase gas price if network is congested

Wrong network error:

Switch to correct network:

Click network dropdown in wallet
Select “Polygon” or “Polygon Amoy”
Approve network switch request
Retry payment

Add network manually if missing:

Network Name: Polygon Amoy
RPC URL: https://rpc-amoy.polygon.technology
Chain ID: 80002
Currency: MATIC
Block Explorer: https://amoy.polygonscan.com

Token approval issues

ERC-20 token approval failed:

Reset token allowance:

// In browser console (advanced users only)
// This resets USDC allowance to zero, then set new amount

Alternative solutions:

Use MATIC instead of ERC-20 tokens
Try smaller approval amounts first
Check token contract address matches network
Verify sufficient token balance

Connection issues

VPN connection failures

WireGuard won’t connect:

Verify configuration:

# Check configuration file syntax
cat rental-config.conf | grep -E "(Address|Endpoint|PublicKey)"

# Test basic connectivity to provider
ping provider-ip-address

Common configuration issues:

Expired configuration (15-minute download window)
Incorrect private key format
Missing endpoint information
Invalid allowed IPs range

Connection drops frequently:

Network stability issues:

Check internet connection stability
Switch from WiFi to ethernet if possible
Disable power saving on network adapter
Try different WireGuard client

Provider-side issues:

Provider internet connection unstable
Provider hardware maintenance
Network routing problems
Firewall blocking VPN traffic

Service access problems

Cannot reach web interfaces:

Verify VPN connection:

# Test tunnel connectivity
ping 10.77.0.1

# Check assigned IP
ip addr show  # Linux/macOS
ipconfig      # Windows

Service status checks:

# Test specific services
curl -I http://10.77.x.2:8888  # Jupyter
curl -I http://10.77.x.2:7860  # Stable Diffusion
telnet 10.77.x.2 22            # SSH

Services not responding:

Wait for deployment:

Initial deployment takes 1-3 minutes
Check deployment status in dashboard
Provider may need time to start services

Alternative access methods:

Try SSH if web interfaces fail
Use direct IP access instead of hostnames
Check for typos in URLs
Verify correct ports

Access method issues

SSH connection problems

SSH connection refused:

# Test basic connectivity
telnet 10.77.x.2 22

# Try verbose SSH for debugging
ssh -v [email protected]

Common causes:

SSH service not started yet
Firewall blocking SSH port
Incorrect username (use rental)
VPN tunnel not established

Authentication failed:

Password authentication:

Use password provided in rental credentials
Check for caps lock or special characters
Try copy/paste to avoid typing errors

Key-based authentication:

# Verify key file permissions
chmod 600 ~/.ssh/gpuflow_key

# Test key authentication
ssh -i ~/.ssh/gpuflow_key [email protected]

# Check if key is loaded in agent
ssh-add -l

Web terminal issues

Terminal won’t load:

Browser compatibility:

Update browser to latest version
Disable ad blockers for gpuflow.app
Enable JavaScript if disabled
Try different browser

Network issues:

Check VPN connection established
Verify terminal URL correct
Try refreshing page
Clear browser cache

Terminal commands not working:

Environment issues:

# Check shell and environment
echo $SHELL
echo $PATH
which python
which nvidia-smi

Permission issues:

# Check user permissions
whoami
groups
ls -la /workspace

VNC desktop problems

VNC won’t connect:

Connection testing:

# Test VNC port accessibility
telnet 10.77.x.2 5901

# Check VNC server status
ssh [email protected] 'ps aux | grep vnc'

VNC client configuration:

Server: 10.77.x.2:5901 or 10.77.x.2::5901
Password: From rental credentials
Color depth: 16-bit for better performance
Compression: Enable for slower connections

Desktop appears frozen:

Resource constraints:

# Check system resources via SSH
ssh [email protected]
htop              # Check CPU/memory usage
nvidia-smi        # Check GPU usage
df -h             # Check disk space

Desktop environment issues:

Right-click desktop for context menu
Open terminal from desktop
Restart desktop environment if needed
Switch to lighter desktop environment

Performance issues

Slow response times

High latency connections:

Network optimization:

# Test latency to provider
ping -c 10 10.77.x.2

# Trace network route
traceroute 10.77.x.2

# Test bandwidth
ssh [email protected] 'iperf3 -s' &
iperf3 -c 10.77.x.2

Connection tuning:

Use wired internet instead of WiFi
Close bandwidth-heavy applications
Choose geographically closer providers
Consider time-of-day network congestion

GPU performance issues:

Monitor GPU utilization:

# Real-time GPU monitoring
watch -n 1 nvidia-smi

# Check GPU memory usage
nvidia-smi --query-gpu=memory.used,memory.total --format=csv

# Monitor temperature and throttling
nvidia-smi --query-gpu=temperature.gpu,clocks.gr,clocks.mem --format=csv

Performance optimization:

# Verify GPU accessibility in Python
import torch
print(f"GPU available: {torch.cuda.is_available()}")
print(f"GPU name: {torch.cuda.get_device_name()}")
print(f"GPU memory: {torch.cuda.get_device_properties(0).total_memory}")

# Check CUDA version compatibility
print(f"CUDA version: {torch.version.cuda}")

Memory issues

Out of memory errors:

GPU memory management:

# Monitor GPU memory in PyTorch
import torch
print(f"Allocated: {torch.cuda.memory_allocated() / 1024**3:.2f} GB")
print(f"Reserved: {torch.cuda.memory_reserved() / 1024**3:.2f} GB")

# Clear GPU cache
torch.cuda.empty_cache()

# Use gradient accumulation for large models
accumulation_steps = 4
for i, batch in enumerate(dataloader):
    loss = model(batch) / accumulation_steps
    loss.backward()
    if (i + 1) % accumulation_steps == 0:
        optimizer.step()
        optimizer.zero_grad()

System memory issues:

# Check system memory usage
free -h
ps aux --sort=-%mem | head -10

# Clear system cache if needed
sudo sync
sudo echo 3 > /proc/sys/vm/drop_caches

Environment-specific issues

Jupyter Lab problems

Jupyter won’t start:

Service management:

# Check Jupyter status
ssh [email protected] 'systemctl --user status jupyter-lab'

# Restart Jupyter
ssh [email protected] 'systemctl --user restart jupyter-lab'

# Check logs
ssh [email protected] 'journalctl --user -u jupyter-lab -f'

Configuration issues:

# Check Jupyter configuration
ssh [email protected] 'jupyter lab --generate-config'
ssh [email protected] 'cat ~/.jupyter/jupyter_lab_config.py'

# Reset Jupyter settings
ssh [email protected] 'rm -rf ~/.jupyter/lab'

Kernel issues:

Kernel not starting:

# List available kernels
ssh [email protected] 'jupyter kernelspec list'

# Install missing kernels
ssh [email protected] 'python -m ipykernel install --user --name python3'

# Test kernel manually
ssh [email protected] 'python -c "import torch; print(torch.cuda.is_available())"'

Package import errors:

# Check Python environment
ssh [email protected] 'which python'
ssh [email protected] 'pip list | grep torch'

# Install missing packages
ssh [email protected] 'pip install package-name'

Stable Diffusion issues

WebUI won’t load:

Service status:

# Check Stable Diffusion WebUI
curl -I http://10.77.x.2:7860

# Check service logs
ssh [email protected] 'systemctl --user status stable-diffusion-webui'
ssh [email protected] 'journalctl --user -u stable-diffusion-webui -f'

Model loading problems:

# Check model files
ssh [email protected] 'ls -la /workspace/stable-diffusion-webui/models/Stable-diffusion/'

# Check disk space
ssh [email protected] 'df -h /workspace'

# Test model loading
ssh [email protected] 'cd /workspace/stable-diffusion-webui && python -c "from modules import shared; print(shared.opts.sd_model_checkpoint)"'

Generation errors:

CUDA out of memory:

# Reduce image resolution
# Use `--medvram` or `--lowvram` flags
# Enable `--xformers` for memory optimization

Model compatibility:

Verify model format (.ckpt, .safetensors)
Check model architecture compatibility
Ensure sufficient VRAM for model size
Try different sampling methods

Mining software issues

Miner won’t start:

Configuration verification:

# Check miner configuration
ssh [email protected] 'cat /opt/miners/t-rex/config.json'

# Test GPU detection
ssh [email protected] '/opt/miners/t-rex/t-rex --list-devices'

# Check pool connectivity
ssh [email protected] 'telnet pool-address 4444'

Permission issues:

# Check file permissions
ssh [email protected] 'ls -la /opt/miners/'

# Fix permissions if needed
ssh [email protected] 'chmod +x /opt/miners/t-rex/t-rex'

Low hashrate performance:

GPU optimization:

# Check current GPU settings
nvidia-smi -q -d CLOCK,POWER

# Monitor during mining
watch -n 1 'nvidia-smi --query-gpu=clocks.gr,clocks.mem,temperature.gpu,power.draw --format=csv'

Mining optimization:

Verify optimal algorithm selection
Check memory clock settings
Monitor temperature and throttling
Ensure adequate power supply

Data transfer issues

Upload/download failures

Large file transfer problems:

Alternative transfer methods:

# Use rsync for resume capability
rsync -avz --partial --progress large_file.zip [email protected]:/workspace/

# Split large files
split -b 1G large_file.zip file_part_
# Transfer parts individually
# Reassemble: cat file_part_* > large_file.zip

Network optimization:

# Compress before transfer
tar -czf archive.tar.gz data_directory/
# Transfer compressed file
# Extract on remote: tar -xzf archive.tar.gz

Transfer interrupted:

Resume transfers:

# Rsync with resume
rsync -avz --partial --progress --append-verify source/ [email protected]:/workspace/

# SCP alternative with resume capability
# Note: SCP doesn't support resume, use rsync instead

Verify integrity:

# Generate checksums before transfer
sha256sum file.zip > file.zip.sha256

# Verify after transfer
ssh [email protected] 'cd /workspace && sha256sum -c file.zip.sha256'

Getting help

Diagnostic information

Before contacting support, collect:

System information:

# Browser information
# - Browser name and version
# - Operating system
# - Extensions installed

# Network information
ping -c 5 gpuflow.app
nslookup gpuflow.app

Rental details:

Rental ID from dashboard
Provider wallet address
GPU model and specifications
Error messages (exact text)
Screenshots of issues

Connection information:

# WireGuard status
sudo wg show

# Network configuration
ip route show
ip addr show

Support channels

Community support:

Discord server - Real-time help
Community forum - Searchable knowledge base

Official support:

Email: [email protected]
Response time: 24-48 hours for technical issues
Include diagnostic information above

Emergency issues:

Payment processed but no access after 10 minutes
Suspected security compromise
Provider completely unresponsive
Data loss during active rental

Self-help resources

Documentation:

Review relevant use case guides
Check provider troubleshooting documentation
Verify wallet and network setup

Community resources:

Search Discord for similar issues
Check GitHub issues and discussions
Review provider community forums

Testing steps:

Try different browser/incognito mode
Test with different wallet if available
Attempt connection from different network
Verify issue persists across multiple rentals
Test with different provider if possible