IET Cluster manual
This is the manual for the usage of the IET cluster. Issues and suggestions can be submitted through the issue tracker.
Getting started
DISCLAIMER: no backup of your data is provided, this is the resposibility of the user!
Support
If you need support regarding the cluster, make sure to consult the support page first!
Account
To get an account for the cluster, please get in touch with the IET support team via iet-support@ost.ch.
Cluster layout
Connect to the Cluster
To connect to the cluster, it is recommended to use ssh
, as it is
available in Linux and Windows (Powershell). For further instructions
on how to connect to the cluster with ssh
, please refer to the ssh
page.
If you prefer a graphical interface check out the vnc page.
Relevant folders
The folders most relevant for general usage of the cluster are given by:
- Your home directory:
shpc0002.ost.ch:/home/OST/<username>
. - The data and project folders, found in
shpc0002.ost.ch:/data
.
For more details about the folder layout refer to the chapter Folder layout.
Software
Software is installed over ansible scripts for proprietary software and via guix for FOSS software.
The list of installed proprietary software can be found in software.yaml.
The list of installed FOSS software can be found in guix-modules.scm.
Modules
To use any specific software on the cluster, you first need to load its corresponding module. To do so:
- List the available modules with
module avail
- Load a module via
module load <module_name>
- Unload a module via
module unload <module_name>
- Purge all loaded modules
module purge
More information can be found in the chapter Environment modules.
Submitting a job
Job management and ressource allocation on the cluster is handled by
SLURM. To submit a
job to SLURM use the slurm-submit
script ssubmit
. It
facilitates general job submission and has also helper commands for
starccm
comsol
matlab
moldex3d
openfoam
Detailed instructions can be found in the README.
More information about SLURM on the cluster can be found in the SLURM chapter.
Job status
To get an Overview of slurm's queue status and the current load on each node, open http://shpc0002.ost.ch in your browser.
To check any job's status, you can:
- Use squeue
to check job-ID, time consumed, and queing status of any job.
- Check the standard output and standard error directed to a file of
the name slurm-%j.out
, where the %j
is replaced with the job
allocation number (found
here).
To cancel a running job, you do scancel <job-ID>
.
Accessing your data locally
To access your data on the cluster locally, you can, for example, use sshfs
, a filesystem client based on SSH.
To mount a filesystem, you type, for example:
sshfs <username>@shpc0003.ost.ch:~/run ~/shpc0003.ost.ch/run
To unmount it, you type the following.
fusermount3 -u mountpoint # Linux
umount mountpoint # OS X, FreeBSD
More information can be found in the chapter Data access.