User documentation for the PLEIADES cluster at the University of Wuppertal
In case of questions, please contact us at firstname.lastname@example.org
Q: Can I get access to the cluster?
A: Yes, if you are part of one of the involved groups. There are also limited resources for any other member of the University of Wuppertal. You can also consult our Quick Reference Card to study the access requirements. To request an account, fill out this form.
Q: How do I apply for an account?
A: Fill out the application form and send it to email@example.com.
Q: How can I use software on the cluster?
A: There is a range of centrally provided software (also see our software documentation). Users and groups typically install their own software additionally in their
Q: Can you install a specific software for me?
A: If the software is interesting to other users/groups and is mentioned on the list of supported software by EasyBuild, we can give it a try. We cannot promise to fulfill any installation or update request, since managing software is a very time consuming task, but we will respond to requests on a best effort basis.
Q: Why is loading software modules so slow?
A: Software modules are shipped via our BeeGFS shared file system. It can happen that running computations produce a heavy load on BeeGFS, which negatively affects the performance.
Q: How can I do file transfers?
A: See our filetransfer documentation.
Q: Where should I compile my programs?
A: Our login nodes have different hardware than our compute nodes (CPUs, InfiniBand). This is why it is a good idea to generally compile where you compute.
Q: Do I have to make sure that I don’t use too many resources?
A: Yes and no. You should always try to minimize core-hours and number of jobs in order to be efficient. But if your computations are a requirement for your research, you can submit the jobs in any case. Slurms fair-share system will automatically ensure a fair scheduling of jobs between users in a group and between all groups.
Q: Why is my job not starting?
squeuecommand has a
Reasoncolumn, providing more information of why pending jobs are not starting. Maybe there are jobs with a higher priority. Some requests are impossible to schedule, because of conflicting requirements, e.g. more than 3 days on the
normalpartition, too many CPU cores per Node (max 64), etc..