SLURM Migration

SCARF is migrating to the SLURM batch system, instructions on this page which refer to LSF commands, parameters or configuration do not apply to SLURM. Please see our SLURM page for more information.

Understanding the SCARF queues

As SCARF has grown over the years it has evolved into a heterogeneous cluster and not a single multi-node machine. As a result, in order to get the best results, it is important to review the resources available on it and to understand how the queues give access to different parts of the cluster.

The scarf queue

The scarf queue is the master queue, and will run jobs on any part of the cluster that is generally available.

  • Jobs with no special requirements
  • 24 hour default run limit, up to 7 days if specified at submission time

The scarf-lexicon, scarf-lexicon-2, scarf-isis-iac and scarf-ibis queues

Over the last several years several groups have gone into partnership with SCARF to purchase dedicated hardware on which to run their codes. These nodes are integrated into the main SCARF cluster, which both simplifies the administration and allows the groups to take advantage of the other features of SCARF, such as data storage and Grid access.

Access to these nodes is only available via these dedicated queues, and only named people have access to the queues. As such, for most SCARF users, these nodes are not available for use.

Other groups are more than welcome to undertake similar partnerships, and are invited to contact us for further details if interested.

Guaranteed Resource pools

SCARF13 and SCARF14 have hardware set aside for use of ISIS and CLF LSF. This hardware is available to other users but only if their job length is 24 hours or less when submitted via the standard queues. Users with permissions to access those hardware pools will use them automatically when submitting via the standard queues. If you think you should have access to those areas, but do not please contact the helpdesk.