Partitions on SHARK

Partitions on SHARK

Important changes and new guidelines to the partition system on 02 Jan 2026

Changes

  • Runtime limits: Common partitions will now have runtime limits.

  • Interactive jobs: For interactive jobs in all common partitions, any runtime request longer than 10 hours will be reset to a maximum of 10 hours.

New Guidelines

Guidelines for Shark jobs to improve scheduling: 

  • Choose the right partition based on job type, resources, and expected duration.

  • Use seff and sacct after jobs to check resource usage and optimize your next job.

  • Request only the resources you need: runtime, memory, CPU, GPU.

  • Do not run interactive sessions longer than 10 hours.

  • Avoid reserving maximum time or resources if you won’t use them.

 

All changes made will be written in the MOTD (Message of the Day), which is displayed when logging in through SSH or by using the command:

$ cat /etc/motd

Changes to gpu partitions on 06 Jan 2026

new partition names

walltime

Nodes

new partition names

walltime

Nodes

gpu

10hrs

res-hpc-gpu[01-02]

gpu-short

30min

If possible ALL gpus from all groups give 30min of runtime to everyone, else all common gpu nodes

gpu-medium

2days

res-hpc-gpu[06-08]

gpu-long

7days

res-hpc-gpu[11-12]

lumcdiagnosticsgpu

inf

res-hpc-gpu[09]

 

List of partitions on SHARK

The following partitions are available on SHARK:

Partition name

Nodes

Default

MemPerCpu

DefaultTime

Max Walltime

Remark

Partition name

Nodes

Default

MemPerCpu

DefaultTime

Max Walltime

Remark

all

res-hpc-exe[001-003,005,007-009,011-014,024,027,029-031,036,039-043],res-hpc-gpu[01-02],res-hpc-mem[01-02]

Yes

2048

1:00:00

-

Default partition

gpu

res-hpc-gpu[01-04,09]

No

2048

1:00:00

-

Only for GPU/CUDA calculations

highmem

res-hpc-mem[01-02]

No

2048

1:00:00

-

For memory intensive applications

highmemgpu

res-hpc-gpu07

No

2048

-

-

For memory intensive GPU/CUDA applications

LKEBgpu

res-hpc-lkeb[04-09]

No

2048

-

-

Only for GPU/CUDA calculations

PATHgpu

res-hpc-path[01-02]

No

2048

-

-

 

RTgpu

res-hpc-gpu[05-06,08,10-11]

No

2048

-

-

 

short

res-hpc-exe[001-003,005,007-009,011-014,024,027,029-036,039-043],res-hpc-gpu[01-06,09],res-hpc-path[01-02]

No

2048

-

1:00:00

max 60 cores for non GPU calculations

A job must always be submitted to a partition (queue). By default, the jobs will be submitted to the "all" partition unless you specify a different queue.