Slurm Partitions

A “partition” is Slurm’s way to refer to a distinct job queue. Each partition has different amounts and types of compute nodes assigned to it, as well as different restrictions on the resources that you can request.

The VACC clusters use the following partitions:

general

This partition is the default partition for Slurm, used if no partition is specified, and assigns jobs to most of the nodes included in the cluster. (Notably, it does not contain nodes with GPUs at this time.) Default walltime is 30 minutes, maximum time is 48 hours.

short

This is the debugging or "short task" partition. It has a default time of 30 minutes, and a maximum time of three hours.

week

This partition is for jobs that need up to one week of walltime to run. It runs on a smaller set of nodes than the general partition.

nvgpu

This is the NVIDIA GPU partition. It has a maximum time of 48 hours and no memory limit. GPU focused jobs which use the NVIDIA GPUs should use this partition. Do not use this partition if your job does not use a GPU.

gpu-debug

This parition contains one node with eight NVIDIA V100 GPUs in it. The partition is meant ONLY for debugging, especially interactive debugging, and should not be used to run jobs. There are strict limits on this partition, though anyone can submit jobs to it.

Only two jobs can be submitted per user at a time
Only one job per user will run at a time
Only one GPU can be requested per job
No more than four CPUs can be requested per job
No more than 75 GB of memory can be requested per job
No more than three hours of run time can be requested per job

gpu-preempt

This partition currently contains two nodes, each with four H200 GPUs. These nodes belong to the CRREL group, but when they are idle, others may run jobs on them. If a user from the owning research group submits a job, and the only available resources are being used by someone from outside the group, outside jobs will be cancelled to make room for the job from the group. Do not submit jobs to this partition if you are not willing to have them cancelled.

hgnodes

This is an NVIDIA H100 partition (SOCKS-funded), and has restricted access. If you are a SOCKS researcher, please contact vacchelp@uvm.edu.

goldenmaple

This partition was purchased by the Cold Regions Research and Engineering Laboratory (CRREL) and has restricted access to this partition. Non-CRREL users have access via the gpu-preempt partition (see above).