Getting Support on CSD3

The University of Cambridge provides support for users on CSD3 and can be contacted via email at support@hpc.cam.ac.uk. This will open a ticket with the support team.

Tip

Each support request should be submitted as a brand new email to the address above. If you reopen a previously closed ticket, your request may be missed because it won’t reach the active service desk agent.

To help us help you, please provide as much information as possible in your ticket. At a minimum, please include:

  • Your name
  • Your account/PI
  • your job submission script
  • the working directories or paths you are using (e.g. /rds/user/$USER/hpc-work/foo/bar/), and
  • as much detail as possible on what you are working on, what you have tried, and error messages you receive.

Use of AI Tools and Online Resources

We understand that as AI assistants such as ChatGPT, Claude, and Gemini increase in popularity, our users may consult them for assistance with technical issues on our platforms. We have compiled in this section some suggestions on how to safely work with these tools as well as how to engage the service desk.

Before acting on any AI-generated advice, please take a moment to review our guidance below on how to use these tools safely and responsibly.

Note

Much of the advice that follows is equally applicable to non-AI generated content found on the internet.

Please consult the following advice:

  • Understand the limitations: AI chatbots do not have the same level of insight or familiarity with CSD3/Dawn as the human support team who maintain these systems.
  • Verify before you act: Do not copy and run commands or instructions provided by AI models without review. Be especially cautious with:
    • Suggestions to use sudo commands or any other privilege escalation routes
    • Software installation steps that may be inappropriate or incompatible with our hardware/software stack(s)
    • Requests for login or sensitive credentials
  • Be security conscious:
    • Be aware of data retention: Most AI assistants/websites retain or process data externally. Anything you share could be stored or used to improve the model; treat prompts as public information.
    • Avoid sharing system-specific details: Do not include configuration files, system paths, network layouts, or architecture diagrams that could expose internal infrastructure details.
    • Do not upload logs or files directly: These may contain sensitive metadata, user data, or internal identifiers that should remain confidential.

Tip

If you need to illustrate a problem, replace identifying information (e.g., usernames, project codes, or directory paths) with placeholders such as <user>, <project>, or <path>.

  • Protect your information: Never include sensitive or identifying data in your prompts. This includes:
    • Account names
    • Passwords
    • SSH/API keys or other sensitive passcodes
    • IP addresses or hostnames
    • Any data protected under institutional, legal or intellectual property confidentiality policies

Attention

If you think you have shared sensitive information: make sure to reset the exposed credentials and let us know by contacting support@hpc.cam.ac.uk. In your e-mail, please provide a description of what happened, including as much relevant information as possible.

What Can I Do?

The following items should help offer some clarity on how to engage with the CSD3/Dawn support team. Please consult and follow this advice.

  1. Communicate clearly with support: when contacting the Service Desk, describe your original issue rather than the diagnosis or solution suggested by an AI assistant. Be sure to include the full context of the problem in your ticket.
  2. Be explicit: use a descriptive subject line. In the body of your email, provide as much relevant information as possible and avoid pasting errors or log files without context. Users should include any error or log files, submission scripts, error messages, and any relevant commands used that would help the support team to retrace your steps and diagnose the fault.
  3. Cite AI use when relevant: if AI suggestions informed your troubleshooting, mention this to Service Desk staff so they can identify and correct any misinformation. You won’t be penalised for telling us you’ve used AI to troubleshoot your problem.
  4. Prioritise official documentation: treat AI advice as a potential aid, not an authority. The official CSD3/Dawn documentation and Service Desk are your primary sources of truth and troubleshooting for all technical issues.
  5. Contact us early: if you’re unsure whether an AI-generated suggestion is safe or reasonable, reach out to the Service Desk before executing it.
  6. Report unsafe suggestions: if you encounter dangerous or misleading instructions from an AI model, please report them so we can raise awareness internally.