Using GitHub Issues for Cluster CEAB Communication

GitHub Issues is the recommended channel for reporting problems, suggesting improvements, and asking questions related to the cluster.

Using GitHub Issues helps us to:

• Centralize communication.
• Track suggestions and their status.
• Avoid duplicated questions.
• Share solutions with the entire user community.
• Build a searchable knowledge base over time.

To report a problem, suggest an improvement, or ask a question, visit the Cluster CEAB GitHub Issues page. You will need a GitHub account to post any issue.

You can create an issue to:

• 🐛 Report a problem: Job failures, software module errors, access problems, storage issues, or unexpected behavior.
• 💡 Suggest an improvement: Propose additional modules, or recommend improvements to services and documentation.
• ❓ Ask a question: Questions about SLURM, storage, software compilation, job execution, resource allocation, or best practices.
• 📚 Request documentation: Identify procedures that require clearer or more detailed instructions.

Before creating a new issue, use the search function to check whether a similar question or problem has already been reported.

1. Open the repository's Issues tab.
2. Select New issue.
3. Write a short and descriptive title.
4. Include all the information needed to understand the request.
5. Check that the issue does not contain sensitive information.
6. Publish the issue.

The title should briefly describe the problem, suggestion, or question.

Good examples:

• OpenMPI job fails when using more than one node
• Question about temporary storage limits
• Documentation request for job submission

Avoid vague titles such as:

• Help
• It does not work
• Cluster problem
• Urgent
• Question

Whenever possible, include:

• A clear description of the problem.
• The approximate date and time of the incident.
• The node, partition, or queue used.
• The job identifier, such as the Job ID.
• The command executed or the submission script.
• The software modules and versions used.
• The complete error message.
• The expected behavior.
• The steps required to reproduce the problem.
• Whether the problem happens consistently or only occasionally.

Example:

snippet.text

Title: OpenMPI error in the compute partition
 
Description:
The job terminates with an MPI communication error when using 32 tasks
across two nodes.
 
System information:
- Job ID: 123456
- Partition: compute
- Nodes: 2
- Tasks: 32
- Module: openmpi/4.1
- Date and time: 2026-08-20, approximately 10:30
 
Steps to reproduce:
1. Load the openmpi/4.1 module.
2. Compile the application using mpicc.
3. Submit the script using sbatch.
4. The error appears when mpirun starts.
 
Error message:
[Paste the complete error message here]
 
Expected behavior:
The application should run using 32 MPI tasks across two nodes.

Format commands and scripts using fenced code blocks. To create one, type three backticks ( ``` ) on a separate line before and after the content.

For a command:

snippet.bash

module load openmpi/4.1
sbatch job.sh

When sharing a large script, provide the smallest version that still reproduces the problem.

When proposing an improvement, explain:

• What change you are proposing.
• What scientific or technical need it would address.
• Which users, research groups, or projects could benefit.
• Whether an alternative is currently available.
• How urgent the request is and why.
• Links to official documentation, when applicable.

When asking a question, describe:

• What you are trying to achieve.
• What documentation you have consulted.
• What commands or configurations you have already tried.
• What result you obtained.
• Which specific part you do not understand or which step is not working.

Example:

snippet.text

Title: How should I request memory for a SLURM job?
 
Goal:
I need to run a Python application that uses approximately 40 GB of memory.
 
What I have tried:
I submitted the job using the following option:
 
#SBATCH --mem=40G
 
Question:
Does --mem=40G request 40 GB for the entire job or 40 GB per CPU?
 
Documentation consulted:
[Add a link to the relevant cluster documentation]

Do not publish the following information in an issue:

• Passwords.
• SSH private keys.
• Access tokens.
• API keys.
• Login credentials.
• Personal information.
• Confidential scientific data.
• Unpublished results that must remain private.
• Configuration files containing secrets.
• Internal network details that are not intended to be public.
• Information covered by confidentiality agreements.

Before publishing logs, scripts, or configuration files:

1. Review the content carefully.
2. Remove usernames if they are considered sensitive.
3. Replace private paths with placeholders.
4. Remove tokens, passwords, keys, and credentials.
5. Check that the output does not contain confidential research data.

For example:

snippet.text

/home/username/confidential-project/results.dat

can be replaced with:

snippet.text

/path/to/project/results.dat

If resolving the issue requires sharing sensitive information, mention this in general terms. The support team will provide an appropriate private communication channel.

After creating the issue:

• Check your GitHub notifications.
• Respond to questions from the technical team.
• Add new information to the same issue.
• Avoid creating duplicate issues for the same problem.
• Use reactions such as 👍 to indicate that you are experiencing the same problem.
• Mention relevant changes, such as a new error message or a different job ID.
• Keep the discussion professional, respectful, and collaborative.

If you find the solution independently, add it to the issue before closing it. This may help other users with the same problem.