Skip to content

How to write good service requests#

Let's Enhance Your Support Experience!

Crafting effective support requests not only helps our team but also benefits you greatly.
With a clear understanding of your issue, we can provide quicker, more precise assistance.
Please take a moment to go through this guide to improve our communication and support efficiency. Thank you!

Before reaching out, please try the search function in our documentation — you might find that your question has already been answered in an existing page or FAQ entry.

Please use your institutional email address, not @gmail,com or @outlook.com

Always use the central support email#

For the best service, please send your requests to hpc-support@fau.de

  • By doing this, your request reaches the entire support team instantly, not just a single person.
  • This allows us to manage and distribute queries efficiently, especially when some team members are unavailable.

One issue, one request#

Service requests are organized by a ticketing system. It groups and categorizes your requests at arrival and assigns a ticket ID.

  • Please submit separate requests for unrelated issues, which helps us assign the right resources, speed up response times, and reduce extra work on our end.
  • Avoid using old and resolved service requests for new issues.

Stay responsive after submission#

We aim to resolve your issues as fast as possible, but we may need further details from you even after submission. Please stay connected and be prepared to provide additional information or run tests as needed.

Review our documentation and conduct quick web searches#

Before contacting support, please review our documentation and perform a quick search online. Often, using the exact error messages and application names can lead you directly to a solution.

Provide a clear explanation of your issue#

Use a descriptive email subject#

A well-thought-out subject line allows us to quickly direct your query to the right expert. Please avoid vague subjects like "It does not work!"

It is always good to mention:

  • Your HPC account (especially if you have multiple accounts).
  • The cluster system where you experience the issue.
  • Name of the application you are using.

Good email subject

GROMACS: I/O errors on Fritz when reading very small files - <account>

Provide as much information about the issue in the support request#

In many cases, the support requests do not provide enough information to the support staff to directly solve the issue. They will ask for information if needed but it is always good to include already as much valuable information in the request as possible.

Although you likely mentioned it already in the subject but subject titles can change, so please provide again:

  • Your HPC account (especially if you have multiple accounts)
  • The cluster system where you experienced the issue
  • Name of the your application. If it is self-developed code, mention the field you are working in (Life Sciences, Computer Science, etc.) and characteristics of the code like programming language, used runtime, etc.

Other valuable information covers:

  • Please indicate whether you're experiencing the issue for the first time or if it was working previously.
  • Are you using one of our provided modules?
    Did you install the application and its dependencies yourself through a package manager (conda, pip, spack, etc.) or from scratch? Have you installed it in a container?
  • Provide job IDs of any jobs showing the issue, and attach relevant batch scripts.
  • Please attach the complete error output as text. If short, paste it directly to the request, otherwise attach it as text file. You can remove application output if unrelated but keep all error and system messages.
  • In case of login issues, check the SSH troubleshooting page and provide ssh -vv <hostname> output.
  • Please let us know if you automatically load any modules or set environment variables in ~/.bashrc, ~/.profile, or ~/.bash_profile.

The support team often needs the commands or outputs to reproduce the issue. Images of outputs slow down the investigation because copy&pasting textual information out of a picture is not that simple. The text outputs (error messages, ...) do not need to look good, we are only interested in its message and not its style.

Please be aware that we might not be able to reproduce your issue for self-built software with many dependencies because it would take too much time to reproduce all the steps you did for setting it up.

Simple case#

Please be specific about the issue, providing all commands executed beforehand. Some problems may only appear under specific conditions, so full details are crucial (modules loaded, commands executes, environment variables set/deleted, etc.). If you got error output, provide the full output attached as a file. The error might happen already in the beginning but the error shows up only later.

Complex cases#

Create a shell script with all the commands you run to replicate the issue. Let us know where input data resides or attach it if feasible. We only access your directories if you explicitly permit it.

Make sure the example is small and runs fast. If it takes a day to compile and another week to run it, it will take a long time until the support team can resolve your issue. Debugging long running applications is challenging. It is likely you can reproduce the issue with a smaller working set and execution environment. Working on it might give you already the right hint and a service request might not be needed anymore.

The XY problem#

Often if a problem X occurs, one tries to find a workaround Y. If this workaround Y is not working as expected, you might want to submit a service request for it. The support staff tries to help resolving the issue with Y but are confused because it is not something they would expect. After long discussion and many email back and forth, the case changes because you actually want to solve problem X, not Y.

Please tell use the fundamental problem X even if you report the issues with Y.

Example

The performance to read in <file format> files with my self-written C application from filesystem $WORK is slow, so I tried using <other file format>. When installing the library for <other file format> from its sources (link), I get an error about missing dependencies to library libfoo.

By following these tips, you'll help us provide the best support experience possible! Thank you!