Skip to content

Latest commit

 

History

History
103 lines (58 loc) · 3.81 KB

faqs.rst

File metadata and controls

103 lines (58 loc) · 3.81 KB
Raw

Frequently Asked Questions

In this section, we'll discuss solutions to common problems. Please work through these suggestions before reaching out for assistance.

Login and Connection Issues

Unable to Log In

If you’re sure there’s no issue with your password or remote-access client but you still cannot log in, you may have exceeded your filestore quota on the cluster. If your $HOME area is full, critical system files may become corrupted, affecting login.

To resolve this:

  1. Attempt to reset your environment with the following command, replacing variables as needed:
ssh -t $USER@bmrc-hpc.shu.ac.uk 'resetenv -f'
  1. If this doesn’t work, and you suspect your $HOME area is full, please contact [email protected].

Problems Connecting with WinSCP

If you experience connection issues with WinSCP, try disabling Optimize Connection Buffer Size:

  1. Open WinSCP and go to Session -> Sites -> SiteManager.
  2. Select your site and click Edit.
  3. In the Advanced Site Settings, under Connection, untick Optimize Connection Buffer Size.

File and Disk Space Issues

Exceeding Disk Space Quota

Each user has a fixed disk space quota in their home directory. Exceeding this quota can cause problems, such as being unable to start applications or submit jobs.

To check if you’ve exceeded your quota, run:

[user@node01 [bmrc-hpc] ~]$ quota -s

If you’ve reached your quota, delete files to free up space. If you can’t log in due to a full quota, contact [email protected].

Windows-Style Line Endings

If you prepare text files (e.g., job scripts) on Windows, they may not work on Unix due to line-ending differences. To convert Windows line endings to Unix, use dos2unix your_file.

Terminal Issues

If commands aren’t working and only bash-4.1$ appears instead of your username, you may have deleted your .bashrc or .bash_profile files, which are hidden files in your home directory.

To restore these files, use resetenv.

Job Submission and Resource Issues

Batch Job Terminates Without Messages or Warnings

If a batch job started with sbatch terminates without messages, it may have exceeded its memory or runtime limits. Ensure you allocate sufficient resources when submitting the job.

Job Stuck in Queue or Pending (PD) for a Long Time

To check why a job is pending, use squeue --reason.

Common reasons include insufficient resources or dependencies on other jobs. Check if you have specified a constraint that limits available nodes.

Interactive Session Memory Issues

By default, an interactive session provides 2GB of RAM. To request more memory, use the srun command with a memory flag, e.g.:

srun --cpus-per-task=2 --mem=8G --pty bash -i

This requests 8GB of RAM (real memory).

Failed Jobs Due to OUT_OF_MEMORY or NODE_FAIL

If jobs fail due to memory issues, ensure the requested memory is sufficient. Use Slurm's --mem flag to allocate memory explicitly.

Software Installation

Using 'sudo' to Install Software on Clusters

Users lack privileges to use sudo to install software. You can install applications in your $HOME directory or a sub-directory thereof. A guide for installation of user software and requesting installation will be provided in due course.