15 articles Research Computing

Cluster: SLURM submission for common GPUs: use “gres” parameter

The setup of GPUs on the Duke Compute Cluster in SLURM (the job scheduler) requires a job to specifically allocate any generic resources that it needs: https://slurm.schedmd.com/gres.html Jobs will not be allocated any generic resources unless specifically requested at job submit time using the –gres option supported by the salloc, sbatch and srun commands. The option…

Research Toolkits: Mounting “Data Commons Scratch Space” on a RAPID VM

(Before beginning the steps below, you need to have both “Data Commons Scratch Space” and a “RAPID” VM setup within the https://rtoolkits.web.duke.edu interface.  See KB0025947 for additional information.) The following will make your “Data Commons Scratch Space” available for access directly on your RAPID VM   Windows VM Login to your VM via RDP as…

Requesting Data Commons storage

Consider the following questions prior to requesting Data Commons storage, as we will need these answers to provision a share: (1)  Do you want the storage to be available for mounting on local workstations or on the Duke Compute Cluster (DCC)? (1a)  Do you want to mount the storage as CIFS (aka “SMB”) or NFS?…

Cluster: Point-of-Contact (POC) List

See the responsibilities of the Point-of-Contact (POC) here. POCs are an important part of the shared governance structure of the Duke Compute Cluster, and they play a role in the efficient administration and security of the system overall. For information regarding POCs that is contained in the KnowledgeBase, use this search link: https://knowledge.oit.duke.edu/?cat=858&s=poc Group name PI,…

Cluster: Membership on hpc-announce@duke.edu list is required for cluster users

In order to have an efficient way to contact all cluster users, all accounts created for cluster users will also add users’ email addresses to the “hpc-announce@duke.edu” email list. This list is used to announce planned maintenance schedules, notify users of persistent performance problems, emergency actions, and the like. Only systems administrators and cluster computing…