By default, jobs submitted to a CycleCloud cluster will request cores from the noderray called “execute”. If a job requires a different nodearray (for example if certain jobs within a workflow have a high memory requirement) you can specify a slot_type attribute for the job. For example, to request a node from the “highmemory” nodearray instead of “execute”, you can add +slot_type = "highmemory" to an HTCondor submission script, or add -l slot_type=highmemory to an GridEngine submission.

On a GridEngine cluster, the job will only run on the specified slot_type and will stay in queue until that a node with that slot_type is available.

On an HTCondor cluster, the job could still be scheduled to run on any available nodes, since the custom slot_type attribute is not included in the job’s requirements or rank expressions. To force the job to only run on that slot type, you can add the slot_type to your requirements expression. Using the “highmemory” slot_type as an example, you can add Requirements = target.slot_type = "highmemory" to your submission script.

For more information, see the Scheduler Integration section of the CycleCloud documentation.