Skip to content

Interactive Sessions

Launch and manage interactive computing environments on CANFAR

🎯 What You'll Learn

  • The differences between session types and when to use each
  • How to launch and connect to sessions quickly
  • How to size resources (RAM/CPU/GPU) appropriately
  • How to manage, share, and troubleshoot sessions

Interactive sessions provide web-based access to powerful computing resources with different interfaces optimized for specific workflows. Whether you're analyzing data in Jupyter notebooks, visualizing radio astronomy images, or running GUI applications, CANFAR's interactive sessions make it easy to get started.

🎯 Session Types Overview

CANFAR supports multiple session types, each optimized for different research workflows:

Session Type Interface Best For Key Features
📓 Notebook JupyterLab Data analysis, coding, documentation Interactive Python, visualization, markdown
🖥️ Desktop Linux desktop GUI applications, legacy software Full desktop environment, X11 apps
📊 CARTA CARTA viewer Radio astronomy visualization Cube analysis, region tools, catalogs
🔥 Firefly Firefly viewer Optical data, table visualization Image viewer, catalog overlay, cutouts
⚙️ Contributed Various Community applications Specialized tools, custom interfaces

🚀 Quick Start Guide

Step 1: Access the Science Portal

  1. Login to CANFAR Portal
  2. Navigate to "Science Portal"
  3. Click the plus sign (+) to create a new session

Step 2: Choose Session Type

Select the interface that best matches your workflow:

Session Type: notebook
Container: astroml
Use Case: Python analysis, Jupyter notebooks, data exploration

Session Type: carta
Container: carta
Use Case: Radio cube visualization, source analysis, imaging

Session Type: desktop
Container: desktop or astroml
Use Case: CASA, DS9, image viewers, legacy tools

Session Type: firefly
Container: firefly
Use Case: Optical data, catalog overlays, image cutouts

Step 3: Configure Resources

Session Name: Choose a descriptive name (e.g., "galaxy-photometry", "alma-reduction")

Choosing Resources

  • Start with defaults (2 cores, 8 GB RAM). Scale up only if needed
  • Large data cubes or catalogs benefit from 32GB+ RAM
  • GPU is only needed for ML or specialized GPU-enabled workflows

Memory (RAM): - 8GB: Light analysis, small datasets - 16GB: Default, suitable for most work - 32GB+: Large datasets, memory-intensive tasks

CPU Cores: - 2 cores: Default, recommended for most tasks - 4+ cores: Parallel processing, intensive computations

GPU (if available): - None: Standard CPU-only work - 1 GPU: Machine learning, image processing

Step 4: Launch and Connect

  1. Click "Launch" and wait for initialization (~30-60 seconds)
  2. Session appears on your portal dashboard
  3. Click the session icon to connect
  4. Start working in your interactive environment

📱 Session Management

Session Lifecycle

stateDiagram-v2
    [*] --> Launching: Click Launch
    Launching --> Running: Session ready
    Running --> Suspended: Close browser
    Suspended --> Running: Reconnect
    Running --> Terminated: Delete session
    Terminated --> [*]

    note right of Running: Access from any device
    note right of Suspended: Session continues running

Session Limits

Limit Value Notes
Concurrent sessions 3 active sessions Across all interactive session types
Session duration 4 days maximum Can be renewed indefinitely
Idle timeout None Sessions run until manually deleted
Storage Persistent Files saved to /arc/ persist

Managing Active Sessions

From the Science Portal

  • View all sessions: Portal dashboard shows active sessions
  • Connect to session: Click session icon
  • Extend session: Use "Renew" button before expiration
  • Delete session: Click "X" to terminate and free resources

From Command Line

# List your active sessions
curl -H "Authorization: Bearer $TOKEN" \
  https://ws-uv.canfar.net/skaha/v0/session

# Delete specific session
curl -X DELETE \
  -H "Authorization: Bearer $TOKEN" \
  https://ws-uv.canfar.net/skaha/v0/session/SESSION_ID

Session Sharing

Share running sessions with collaborators in your group:

  1. Copy session URL from browser address bar
  2. Share with team member (must be in same CANFAR group)
  3. Collaborate in real-time - both can see and modify the same session

Security Note

Only share session URLs with trusted collaborators. Anyone with the URL and proper group membership can access your session.

🔧 Advanced Session Features

Resource Allocation

Memory Management

# Check memory usage in session
free -h
htop

# Monitor specific process
ps aux | grep python

Memory Tips: - Start with default 16GB, increase if needed - Large datasets may require 32GB+ - Memory is shared - be considerate of other users

CPU Usage

# Check CPU cores available
nproc

# Monitor CPU usage
top
htop

# Run parallel processing
python -c "import multiprocessing; print(f'CPUs: {multiprocessing.cpu_count()}')"

GPU Access

For containers with GPU support (e.g., astroml-cuda):

# Check GPU availability
nvidia-smi

# Test GPU in Python
python -c "import torch; print(f'CUDA available: {torch.cuda.is_available()}')"

# Monitor GPU usage
watch -n 1 nvidia-smi

Persistent Configuration

Save personal settings that persist across sessions:

# Jupyter configuration
mkdir -p /arc/home/$USER/.jupyter
cp jupyter_config.py /arc/home/$USER/.jupyter/

# Shell configuration
echo "alias ll='ls -la'" >> /arc/home/$USER/.bashrc

# Python packages (user installation)
pip install --user astroplan # in some containers, the --user may not be needed

Session Networking

Port Forwarding

For custom web applications running in sessions:

# In session: Run application on specific port
python -m http.server 8080

# Application accessible at: https://SESSION_URL/proxy/8080/

External Access

Sessions are accessible from anywhere with proper authentication:

  • Same computer: Original browser window
  • Different computer: Copy session URL, login with CADC credentials
  • Mobile device: Session URL works in mobile browsers

🛡️ Security and Best Practices

Data Security

✅ Do: - Save important work to /arc/projects/ or /arc/home/ - Use group permissions for collaborative data - Regularly save and backup critical results - Log out of shared computers

❌ Don't: - Store sensitive data in /scratch/ (wiped at session end) - Share session URLs publicly - Leave sessions running unnecessarily - Store passwords in plain text files

Performance Optimization

Session Performance

# Close unused applications to free memory
# Kill runaway processes
kill -9 PID

# Clean temporary files
rm -rf /scratch/temp_*

Storage Performance

# Use /scratch/ for intensive I/O
cp /arc/projects/data.fits /scratch/
# ... process in /scratch/ ...
cp /scratch/results.fits /arc/projects/

# Compress large files
gzip large_dataset.fits

Troubleshooting Common Issues

Session Won't Start

Problem: Session stuck in "Launching" state

Solutions: 1. Check resource availability - try lower memory/CPU 2. Wait 2-3 minutes for container download 3. Try different container image 4. Contact support if persistent

Session Disconnected

Problem: Lost connection to running session

Solutions: 1. Refresh browser page 2. Click session icon again from portal 3. Check internet connection 4. Clear browser cache if needed

Out of Memory Errors

Problem: Application crashes with memory errors

Solutions: 1. Launch new session with more memory 2. Process data in smaller chunks 3. Use /scratch/ for temporary files 4. Optimize code for memory efficiency

Slow Performance

Problem: Session responding slowly

Solutions: 1. Check system resources (htop, free -h) 2. Close unnecessary applications 3. Use /scratch/ for I/O intensive tasks 4. Consider launching session with more resources

🔗 Session-Specific Guides

Detailed guides for each session type:

🔗 What's Next?

Once you're comfortable with interactive sessions:

Session Success Tips

  1. Start small - Use default resources and scale up if needed
  2. Save frequently - Important work should go in /arc/ directories
  3. Share wisely - Session URLs are powerful - only share with trusted collaborators
  4. Monitor resources - Keep an eye on memory and CPU usage with htop
  5. Clean up - Delete finished sessions to free resources for others