Using CIR systems from MS Windows

Introduction

The central Computationally Intensive Research (CIR) / HPC facilities such as CSF, DPSF/Hydra, Redqueen, Zrek and iCSF are Linux systems but can easily be accessed from your Windows desktop. Here we provide some useful information for those users accessing the CSF, DPSF etc. from Windows and who are less familiar with Linux than Windows.

Logging in from Windows

We now provide common access instructions for all of the central facilities (CSF, DPSF/Hydra, Redqueen, iCSF, zCSF). Please follow the connecting from Windows instructions.

Important information about batch submission scripts

If you use an editor on Windows (e.g., Notepad) to create a batch job submission script and transfer it to the CSF, DPSF or Zrek, your batch job will error and fail to run. This is because Windows and Linux have different ways of indicating end-of-lines in text files (using hidden characters) and the batch system cannot understand the Windows line-end convention. We have three recommended ways of dealing with this:

  1. Use gedit on the CSF, DPSF or Zrek login node to write your jobscript – although this is a Linux editor it is GUI based and can be run directly on the CSF, DPSF or Zrek which means you do not need to transfer files back and forth to your desktop. If you’ve logged in with MobaXterm then this will work straight away. If you’ve used PuTTy you must also run XMing. If you’ve logged in from a Mac then you’ll need to run XQuartz (all of which provide X11 on your local computer).
  2. If you do create your jobscript text file under windows and upload it to the CSF, DPSF or Zrek, run the following command on the login node before you submit it to batch:
    dos2unix filename

    where filename is the name of your file. Please note: you should never run dos2unix on other files you may have uploaded to the CSF, DPSF or Zrek from Windows. It should only be used to convert text-file jobscripts. If you run dos2unix on other files such as data files you will break them – even if you used a Windows application to create the data files. dos2unix is for converting jobscripts only.

  3. On Windows install and use Notepad++. Within Notepad++ select Edit menu -> EOL Conversion -> UNIX/OSX Format when creating/editing a new text file. When you then upload that file the CSF will understand it.

Our recommendation is to use gedit (or any other Linux editor) directly on the CSF, DPSF or Zrek.

File and directory/folder names

The use of spaces and non-alpha-numeric characters in filenames and directories on Linux can cause problems. In particular, the batch system may not be able to fully resolve the names of directories and cause jobs to error. Command line listing of files will be less clear, and without the use of escape characters navigation of the filesystem, manipulation and deletion of files can be difficult. Acceptable non-alpha-numeric characters are underscore and dash. For example instead of:

Chemistry Simulation 1              # Bad choice of name - it contains spaces

use:

Chemistry_Simulation_1              # A better choice - it contains _ instead of space

or

Chemistry-Simulation-1              # A equally good choice - it contains - instead of space

You may need to ensure that the folder structure and filenames on your windows PC/laptop follows these conventions to avoid problems when you upload then to the CSF. On windows you should not encounter any problems using these conventions, provided you name only your files and folders in that way, i.e. do not change any of the system files or directories on your windows PC/laptop.

Basic Linux Commands and Compute Cluster Usage

If you are unfamiliar with Linux/Unix it is recommended that you work through one of the following:

Approximately three times a year an Intro to CSF course is run which provides an excellent starting point for users new to compute clusters and parallel computing concepts.

Last modified on September 27, 2021 at 7:49 am by George Leaver