Getting Started

Logging onto Boreal

To log on Boreal remotely, use:

ssh –X your_userid@boreal.emsl.pnl.gov

The -X option allows remote users to access X Windows functionality across the internet.

Warning

Due to security concerns with older cipher and MAC (message authentication code) methods used by ssh, the Boreal configuration has been changed to remove many of the older algorithms.

Currently supported Ciphers include:

  • aes128-ctr

  • aes192-ctr

  • aes256-ctr

Currently supported MACs include:

  • hmac-sha1

  • hmac-sha2-256

  • hmac-sha2-512

If you experience problems with ssh connecting to Boreal, please make sure that your ssh client software has been updated to a recent version that in cludes at least some of the above cipher and MAC options. If you continue to have problems, please contact us at mscf-consulting@emsl.pnl.gov

Running Jobs on Boreal

Boreal is a batch scheduled environment managed by Slurm. For commonly used software we have pre-formed submission scripts that either interactively query you for the required input fields or take a prescribed set of arguments. These scripts then build the submission script and submit it for you.

To list the available submit scripts type:

ls /home/scicons/boreal/bin/submit*

Most of these scripts accept a -h or --help as an argument to give a description of how to use them. When using the preformed submit scripts, please be sure to run them from your /dtemp project directory. This is so that output files land in the lustre file system which has much higher bandwidth and lower latency than the /home file system. It is also a good idea to have the input files (if needed) in the /dtemp file system.

Running jobs are subject to the limits set within our Job Policy.

Project Account Names on Boreal

All of the above pre-formed scripts and when you build your own you need to specify which project account is to be used. To see which project accounts you are permitted to run under, type:

gbalance

This will produce a list of the projects you may run under as well as the node hours used and remaining in the project. In the non-header lines of the output, the characters up to the first blank are the project name. If you are building your own submission scripts you will need to define the project in your scheduler directives:

#SLURM --account 'project'

Writing a Batch Script

If you are creating your own submission script, you submit them with the sbatch command:

sbatch 'my_submission_script'

The scheduler will return the SLURM job ID or some error message if your account is incorrect or the scheduler is not accepting jobs. When submitting jobs, please run them from your /dtemp project directory.

More information is available at Job Scripts.