The Farm uses one's JLab CUE account. After you obtain a CUE account, contact your Hall/project computing coordinator to add you to a (POSIX) group that provides Farm access, and to avoid
Invalid account or account/partition combination specified
when submitting jobs (whether with sbatch
/srun
, SWIF, or JupyterHub), your computing coordinator must also assign your user account to one or more batch system (Slurm) accounts, by either filing a ticket on your behalf or performing the association personally.
To access the Farm, SSH using your personal account with 2-factor authentication to login.jlab.org.
and then use your personal or group account with your CUE password or SSH key to SSH to ifarm
. See KB0015066 for more information and examples. OpenSSH can be configured, as can PuTTY, to re-use connections to avoid needing to repeatedly re-authenticate.
If your coordinator has added you to a group and you still can't log in to ifarm
, submit an incident. If you can't log in to jlabl1
, jlabl2
, or jlabl[3-5]
either, or you can't log in to login.jlab.org.
, starting off with that will help your ticket get directed to the right person sooner.
ifarm
is a logical interactive node which will automatically, based on round robin scheme, log you in to one of the following physical interactive nodes:
ifarm2401
, ifarm2402
/scratch
You are free to use these nodes for software development and testing, but do not run long tasks. Tasks may be killed as needed, or the entire host rebooted to correct problems. Instead, submit long tasks as batch jobs using sbatch
or SWIF. Access to off-site nodes is restricted. Request additions to the allow-list, or download to shared storage (e.g. your CUE home directory or /work
) from jlabl[1-5]
. For additional details refer to Farm Cluster Layout.