Below is some documentation which might be helpful to Euler users
Euler uses Slurm as a job scheduler. Users should launch computations using batch scripts via
sbatch wherever possible.
New To Slurm?
Euler-specific Slurm Features
As of April 2019, Euler provides hooks for a shell startup file specific to jobs run in Slurm. Located at
~/.slurmrc, the script allows users to specify commands, environment variables, and modules that they wish to be set for Slurm jobs, but not necessarily for all of their shell sessions.
The use of .slurmrc is optional, but enabled by default on new accounts for users of the Bash shell. Accounts which were created before April 2019 can opt-in to this new functionality by running 1
slurmrc_opt_in at a shell prompt.
Temporary Folders and Network Scratch
Euler makes a small amount of space available for temporary folders on each node under the
/tmp directory. Slurm jobs may reference the
$TMPDIR variable as a job specific folder for temporary files. These folders are automatically destroyed when a job terminates (cleanly), and are not typically available across multiple nodes.
For slightly larger files, or those which might be referenced from multiple nodes, Euler also provides access to an SSD-backed network file system to be used as a scratch partition. Slurm jobs may reference the
$SCRATCH variable for a folder available on any node within a job allocation. These folders are automatically destroyed when a job terminates (cleanly).
As of October 2019, .slurmrc is loaded by default for Slurm jobs executed using bash or fish as the interpreter. ↩