Support and troubleshooting

A practical learning route

Turn errors into reusable knowledge

HPC learning becomes much clearer when the error, job details, environment, and expected result are captured together. Follow this short route to understand a problem before asking for local account guidance.

Guided practiceStart with the evidence

Learning habitKeep one small reproducible example

Locate the failure

Write down the exact command or job step that failed, together with the time it occurred and the result you expected.

Capture the evidence

Keep the job ID, full error message, relevant output files, and tool or environment version. Do not paraphrase an error when you can paste it.

Send a focused request

Explain what you tried, what happened, and what you need next. Remove passwords, access tokens, and restricted research data.

Capture the essentials

For a Slurm job, begin with the scheduler record and the final lines of both output files. Replace the placeholder job ID and filenames with your own values.

squeue --job JOB_ID
sacct --jobs JOB_ID --format=JobID,State,ExitCode,Elapsed,MaxRSS
tail -n 40 job-output.out
tail -n 40 job-error.err

For an environment or software problem, capture the executable and version after loading the modules or activating Conda.

module list
conda info
which TOOL_NAME
TOOL_NAME --version

Use this learning checklist

Copy this compact template into your study notes. It helps you reproduce a problem without guessing about the command, system state, or intended output.

Topic: Short description of the issue

Goal: What I am trying to run.
Where it fails: Command or workflow step.
Expected result: What I expected to happen.
Observed result: Exact error or scheduler state.
Job ID: If applicable.
Environment: Modules, Conda environment, or tool version.
What I already tried: Brief list of troubleshooting steps.
Attachments: Relevant job script and small log excerpts only.

Local guidance

Know what to ask for

Use the checklist above to describe the issue clearly. For local accounts, access, or allocation policies, follow the process supplied for your own system.

Local access questionsContact the administrator of University responsible for your system.

Choose the right learning focus

Connection practiceReview SSH keys, profile syntax, and safe authentication habits.

Command practiceReview paths, modules, and basic cluster commands with short examples.

Slurm practiceInspect resource requests, job state, and log files after a small test run.

Conda practiceCheck activation, installed tools, and a reproducible environment description.

Continue learning

Build a focused HPC study routine

Use these reputable external lessons to deepen your command-line foundations, scheduler understanding, and reproducibility habits. Local access and account rules always come from your own system.

Foundation lessonSoftware Carpentry: The Unix Shell Build confidence with navigation, files, permissions, and small shell scripts. HPC workflowHPC Carpentry: Using the Shell Learn how command-line skills connect to shared research-computing workflows. Environment managementConda: Getting Started Learn to create isolated environments, inspect them, and install packages reproducibly. Scheduler referenceSlurm: User Documentation Explore job states, commands, resource requests, and scheduler concepts from the official reference.

Safe sharing rule: never include passwords, private SSH keys, access tokens, participant information, or restricted research data in a support request.