Instructor Notes
OnDemand Portal Orientation - Instructor Notes
Workshop Overview
This 2.5-hour workshop introduces beginners to the OnDemand web portal for HPC access. Participants learn to navigate the interface, manage files, access the command line, and run interactive applications.
Learning Objectives
By the end of this workshop, learners will: - Successfully log into OnDemand - Navigate the dashboard and find key features - Upload and download files - Access the web terminal - Launch Jupyter notebooks - Submit batch jobs - Monitor job progress
Pre-Workshop Preparation
- Verify all participants have:
- Valid Pomona accounts
- DUO MFA configured
- Active HPC accounts
- Working web browsers
- Test OnDemand access yourself
- Prepare example files for demonstrations
- Have backup plans if cluster is offline
Episode-by-Episode Teaching Guide
Episode 1: What is OnDemand? (15 min teaching + 5 min exercise)
Key Points to Emphasize: - OnDemand is a web interface, not a replacement for SSH - No command-line knowledge required - Graphical file manager is powerful for beginners - Can run multiple applications simultaneously
Demonstration: - Show login screen and dashboard - Explain each menu section - Highlight differences from SSH
Interactive Exercise: - Have learners log in and navigate to Home Directory - Show how to upload a test file - Celebrate first interaction with OnDemand
Common Questions: - “Is OnDemand slower than SSH?” (Slightly, but user-friendly) - “Can I still use SSH?” (Yes, both work together) - “What if I mess up files?” (Most stored safely with backups)
Episode 2: Dashboard Navigation (20 min teaching + 10 min exercise)
Key Points to Emphasize: - Consistent layout across all OnDemand pages - Three main sections: Files, Jobs, Apps - Help is always available via ? icon - Status indicators use color coding
Demonstration: - Walk through each menu item slowly - Show how to access profile/account settings - Demonstrate session management - Explain timeout and re-authentication
Interactive Exercise: - Have learners navigate to each main section - Locate their disk quota - Find current cluster status - Identify available partitions
Tips for Teaching: - Use large screen/projector for visibility - Pause after each section for questions - Have ready answers for “Where do I find…?” - Remind about browser cache if pages look odd
Episode 3: File Manager (20 min teaching + 15 min exercise)
Key Points to Emphasize: - Four main storage locations: /rhome, /bigdata, /scratch, /tmpfs - /rhome and /bigdata share a single 1 TB lab quota on BeeGFS - /bigdata is for collaborative research data - /scratch is non-persistent SSD; auto-deleted when the job completes — there is no grace period.
Demonstration: - Upload a file (preferably <10 MB for speed) - Download it back - Create a directory and move a file - Show quota usage in Accounts/Usage - Demonstrate file permissions
Interactive Exercise: - Each learner uploads a test file - Creates a new directory - Moves file into directory - Downloads file back to verify round-trip - Checks their disk quota
Trouble Spots: - Upload may be slow with many concurrent uploads - Drag-and-drop doesn’t work in all browsers - Some learners may accidentally delete files - Path confusion (/bigdata vs /rhome)
Solutions: - Break class into small groups for uploads - Show explicit Upload button as fallback - Explain deletion is permanent (no trash) - Clarify path with absolute paths in challenges
Episode 4: Shell Access (20 min teaching + 15 min exercise)
Key Points to Emphasize: - Web Terminal is a full bash shell - Same commands as SSH work here - Good for running jobs and file operations - Less suitable for interactive programs (vim, top)
Demonstration: - Launch web terminal from OnDemand - Show basic commands: ls, pwd, cd - Submit a simple SLURM job - Show job output - Demonstrate module loading
Interactive Exercise: - Launch web terminal - Navigate to home directory - Create a new directory - View files with ls - Submit a simple job (sleep 30) - Check job status - View job output when complete
Teaching Tips: - Many learners have never used bash before - Use simple commands first (ls, pwd) - Explain that “command not found” is normal (need module load) - Show job submission without complex SLURM script details
Potential Issues: - Learners may type passwords or sensitive data in terminal - Slow typing on web terminal causes lag - Jobs fail due to typos or missing modules - Terminal disconnections on slow connections
Episode 5: Interactive Applications (20 min teaching + 15 min exercise)
Key Points to Emphasize: - Jupyter notebooks combine code and documentation - RStudio provides full R IDE - Resource allocation determines launch time - Sessions timeout; download work before expiry
Demonstration: - Launch Jupyter notebook - Show cells, markdown, output - Create simple Python code cell - Create matplotlib visualization - Save notebook - Close notebook and show it’s gone (session ended)
Alternatively, show RStudio: - Show script editor, console, environment, plots pane - Write simple R code - Create ggplot visualization - Show package installation
Interactive Exercise: - Each learner launches Jupyter - Creates a notebook with: * import statement * data creation (numpy array or pandas DataFrame) * simple calculation * visualization - Download the notebook - Close the session
For GPU Learners (if available): - Show GPU allocation in launcher - Verify GPU with pytorch/tensorflow - Do simple GPU operation - Monitor with nvidia-smi
Common Challenges: - Long launch times on busy clusters - Not enough GPU resources available - Learners don’t save before session timeout - Package installation errors
Solutions: - Warn about 5-10 minute launch time - Have CPU-only notebooks as backup - Emphasize downloading notebooks - Have pre-installed common packages - Show error messages when they occur
Episode 6: Job Management (20 min teaching + 15 min exercise)
Key Points to Emphasize: - Job Composer is easier than manual scripts - SLURM scheduler manages queue fairly - Jobs can wait hours; patience required - Check output files after job completion
Demonstration: - Create simple job in Job Composer - Submit to normal partition - Show job in Active Jobs view - Wait for job to start (may take 2-10 min) - Show job progressing - When done, view output - Explain stderr vs stdout
Demonstration of Cancellation: - Submit another job - Before it starts, cancel it - Show it disappears or changes status
Interactive Exercise: - Create job that: * Prints start time * Runs calculation (Python, R, or sleep) * Prints end time - Submit job - Monitor until completion - View output - Calculate actual runtime
For GPU Interest: - Show GPU job submission with
--gres=gpu:1 - Explain wait times for GPU are longer - Show
GPU-specific output
Teaching Notes: - Cluster timing may make jobs run immediately or queue - Use this as teaching moment about SLURM scheduling - Some jobs may fail; use as debugging opportunity - Explain error messages clearly
Timing Guide
Handling Common Issues During Teaching
Cluster is Offline/Slow
- Pre-record demonstrations
- Show screenshots if live demo fails
- Explain maintenance windows
- Provide hands-on instructions for later
Many Concurrent Uploads
- Stagger upload exercises
- Break into smaller groups
- Use /scratch for testing (faster)
- Explain bandwidth limitations
Learners Stuck on Login
- Verify Pomona credentials
- Check DUO is configured
- Try different browser
- Have ITS contact ready
Assessment and Feedback
Formative Assessment
During workshop, ask: - “Can you find the Upload button?” - “Which storage location should we use for this 50 GB dataset?” - “What partition would we use for GPU work?”
Resources
Emergency Contacts
- HPC Support: its-hpc@pomona.edu
- ITS Support: (for account issues)
- OnDemand errors: Screenshot and email to its-hpc@pomona.edu
Helpful References
- OnDemand Help: Built-in ? icon
- Sagehen HPC Docs: https://www.pomona.edu/its/
- SLURM Docs: https://slurm.schedmd.com/
- Carpentries HPC: https://carpentries-incubator.github.io/hpc-intro/
Post-Workshop Follow-up
Send learners: 1. Certificate of completion (if applicable) 2. Link to OnDemand help documentation 3. Info about next workshops (Security, GPU, ML) 4. Contact info for support 5. Optional: recorded session video (if recorded)
Notes for Future Instructors
- Include more examples of real research workflows if possible
- Cluster performance varies; plan for waits
- Some learners take time with new interfaces; patience helps
- Celebrating first successful job submission is powerful
- Many learners will want to return for advanced topics
- Consider recording for asynchronous learners
Good luck with your workshop!