RUNNING TRANSP ON THE FLUX CLUSTER (PPPL)

Users must have:

• A valid UNIX login account at PPPL
• A valid TRANSP account
• SSH access to the FLUX cluster

Log in to FLUX:

ssh flux.pppl.gov

After logging in, load the NTCC module required for TRANSP:

module load mod_ntcc ntcc/gcc/13.2.0/24.8.0_flux

Note:
As of today, ntcc/gcc/13.2.0/24.8.0_flux is the only module that supports running TRANSP on the FLUX cluster.

When a TRANSP job is started on FLUX, the system automatically detects the FLUX environment and submits the job to the appropriate partition.

• Default runs are submitted to the long partition
• Use the fast option for short test runs
• Use the big option for long, low-priority runs

To avoid interactive prompts from tr_send, users are strongly encouraged to set:

export MDS_TRANSP_SERVER=TRANSPGRID.PPPL.GOV
export TR_EMAIL=[USERNAME]@pppl.gov

• Job priority on the FLUX cluster differs from the former McCune cluster
• Users are encouraged to explicitly set walltime
• Use the fast or big options when appropriate

Web interface:

https://transpgridmonitor.pppl.gov

Command line (on FLUX login node):

squeue -u "$USER"

While a job is running, the TRANSP log file is available at:

/scratch/shared/$(whoami)/transp_compute/<TOK>/<RUNID>/<RUNID>tr.log

Replace:

• <TOK> with the tokamak name
• <RUNID> with the TRANSP run ID

• If a job fails, the user will receive an email notification
• For PPPL users, completed runs are stored in the MDSplus tree on the TRANSPGRID server at PPPL

1. Start a TRANSP run using the fast partition and interactive walltime selection:

   tr_start <RUNID> fast walltime

   Follow the interactive prompts to complete the setup.
2. Submit the run to the queue:

   tr_send <RUNID>