workflow remote reserve¶

Synopsis¶

workflow remote reserve <remote> <profile> [options...]

Description¶

Submits one or more long-running Slurm compute allocation worker jobs on the specified remote cluster.

Compute workers connect back to the central manager server over reverse SSH tunnels, pulling and executing simulation SPoints across allocated GPU/CPU resources without requiring repeated Slurm job queue wait times.

Arguments¶

<remote>: Name of the registered remote cluster (configured with Slurm).
<profile>: Name of the Slurm profile (defined in sbatch_profiles/<profile_name>.yaml on the remote).

Options¶

-n count, --num-workers count: Number of parallel compute worker allocations to submit to Slurm. Defaults to 1.
--sbatch-args arguments: Additional command-line arguments to append to the Slurm sbatch invocation.
-s stages, --stage stages: Limit worker tasks to specific stages.
-w sweeps, --sweep sweeps: Limit worker tasks to specific sweeps.
--exclude-stage stages, -xs stages: Exclude specific stages from being executed by this worker reservation.
--exclude-sweep sweeps, -xw sweeps: Exclude specific sweeps from being executed by this worker reservation.