Tips
Training Materials
- If you are a new user, or you have not used the system for a while, please take some time to work through the Getting Started section of the documentation, in particular the batch tutorial.
- We also have a training course. Unfortunately there are no dates at the moment due to upgrade work commitments, but it is possible to work through the materials on your own.
- If you are new to using the Linux command line please consult the Linux commands summary page.
Login nodes
- These are used by everyone and are intended for creating batch jobs and basic file management.
- Please do not run any computational processes on the login nodes.
- All computational work must be submitted to batch.
- Software installs and dataset downloads should be done in an interactive batch session (qrsh).
- If you need to copy large files, or lots of files between CSF file systems (e.g. home to scratch, scratch to RDS (typically starting
/mnt
)) please do so via batch jobs. - Did you know you can quickly tarball and compress files via batch using pigz?
- Disruptive processes will be terminated without warning. If time permits we may contact you with some advice.
- Persistent misuse of the login nodes may result in your account being disabled.
Running your work/Batch Jobs
- Please do the 10 minute tutorial frist.
- GPU jobs will not run unless you have asked for access to GPUs.
- Information about queue wait times.
Filesystems
A summary is below, for more details please consult the Filesystems section of this website. Also see the login node section above for some handy hints on dealing with files.
Home Space
- Home file space is not very big. It is primarily for enabling your access to the service.
- If home fills up it can disrupt everyone in your group, in some cases that may be 100 people or more.
- When home fills up the CSF team lose valuable time from their day tracking down who is making excessive use of it and contacting them.
- You should be using scratch to run your jobs.
Scratch
- Intended for temporary files only. There is no backup.
- Important files should be moved to Research Data Storage (RDS, speak to you supervisor/PI to find out if they have any).
- Scratch is shared by all users, please be considerate of this and keep your usage as small as possible.
- Please regularly check your usage with the
scrusage
command - Housekeeping of your scratch files should be done as part of your batch jobs/workflow, please do not wait for our automated 3 month deletion policy to do it for you.
Research Data Storage
- The University recommends that all important data be stored in Research Data Storage (RDS) – this is additional storage visible on the CSF.
- There are two flavours of RDS.
- The one that has a path similar to
/mnt/eps01-rds/MyProject
(theeps
faculty / department code will differ) is available on the CSF. - The one that has a path similar to
\\nasr.man.ac.uk\epsrss$\snapped\replicated\MyProject
cannot be accessed on the CSF.
- The one that has a path similar to
- Make sure you/your PI/Supervisor requests that your RDS should be accessible on CSF, to ensure the correct type of storage is allocated to their project.
Software
- We have a lot of software on the system.
- We cannot document all of it. Please read the software summary webpage for information on how to look for what is available and for tips on how to do your own installs.
- Please note that if you need a centrally installed piece of software updated, various factors can affect how long it will take. Two weeks or more is not unusual.
Getting help, and helping the CSF Team to help you
Please ensure you have consulted the documentation first. We appreciate that some of it is quite dense and may be daunting, but we have tried to be comprehensive and it has been developed as a result of previous enquiries to us.
There is a search function – the box at the top of the menu on the left hand side.
If you cannot resolve your issue please fill in our Connect Portal HPC Help form
- this comes directly to the team
- using the form helps us, in several ways, with workload management
- we have someone on duty Monday-Fri during business hours to handle the enquiries
- provide as much detail as possible
- please submit your request only once to avoid unnecessary duplication and confusion
- please do not email individual members of staff directly as it may not be their day for handling enquiries
We will get back to you in as timely a manner as possible. Enquires where someone is clearly experiencing a serious issue will be prioritised.
Feb 2025: As you will be aware we are currently carrying out upgrades to the service alongside all our day to day operations. During this period you may experience slower response time than usual.
We are not able to help with email issues, printer issues, password issues, timesheet problems, pure enquiries, or anything else that is not listed on https://ri.itservices.manchester.ac.uk/
The main IT ServiceDesk will be able to advise and put you in touch with the right team.