Skip to content

Nodes available for testing the new operating system

Early testing!

Please note that the advice and commands on this page are still being tested. Expect the possibility for bugs and errors, and always test for accuracy by comparing against commands you know already work.

Check back for updates!

This page will continue to be updated with new information. Check back regularly for updates.

ICER will progressively transition nodes to the new operating system.

Development nodes

At the moment, the following development nodes are available for testing access:

  • dev-intel14
  • dev-intel14-k20
  • dev-intel16-ubuntu
  • dev-intel18-ubuntu
  • dev-amd20-ubuntu
  • dev-amd20-v100-ubuntu (NVIDIA v100 GPU)

You can access these via SSH like any other development node:

1
2
ssh user@hpcc.msu.edu  # First SSH into the gateway
ssh dev-intel16-ubuntu  # Then SSH into the desired dev node

After June 17th, these nodes will be accessible without the -ubuntu suffix, i.e.,

  • dev-intel14
  • dev-intel14-k20
  • dev-intel16
  • dev-intel18
  • dev-amd20
  • dev-amd20-v100 (NVIDIA v100 GPU)

SLURM jobs

SLURM jobs need to be submitted from compatible nodes

All SLURM jobs submitted to Ubuntu nodes must be submitted from development node running Ubuntu. The same goes for jobs running the current operating system, so you must use an existing dev nodes to submit jobs to the main cluster.

Additionally, a limited number of compute nodes are available with the new operating system. All jobs submitted to these nodes need to be submitted from one of the Ubuntu development nodes listed above. To run jobs on these nodes, include the line #SBATCH --reservation=ubuntu_compute to a job script with any other resource requests, for example:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
#!/bin/bash --login
#SBATCH --ntasks=1
#SBATCH --mem=20M
#SBATCH --time=01:00:00
#SBATCH --reservation=ubuntu_compute

cat /etc/os-release  # Should show Ubuntu

# Print resource information
scontrol show job $SLURM_JOB_ID
js -j $SLURM_JOB_ID

At the moment, the following clusters are available via SLURM:

  • intel14 (reserved for OnDemand jobs, which are still a work in progress)
  • intel16
  • intel18
  • amd20
  • amd22
  • Data machine (using --account=data-machine)