Concepts
Four mental models that show up across the Jobs section — whether you're working with scripts, templates, submissions, or the catalog.
The pipeline
Jobs flow through three stages. A Template is your reusable blueprint — it defines what to run with placeholders for things like dataset paths or cluster config. When you render a template, Vantage fills in those placeholders and produces a concrete Script — a Slurm batch file ready for submission. Submit that script to a cluster, and it becomes a Submission — a running (or queued, or completed) workload.
Each stage links back to its parent: a rendered Script knows which Template it came from, and a Submission knows which Script it was created from.
Script files
A Job Script has one entrypoint — a .py or .sh file with #SBATCH directives and executable code — plus any number of support files (configs, Python modules, input data) the entrypoint references at runtime.
| File type | Description |
|---|---|
| Entrypoint | .py or .sh — the file Slurm executes |
| Support | Any auxiliary file the entrypoint needs |
Archive vs Delete
Archive hides it, delete wipes it.
Archive removes a resource from your default view but keeps everything intact — data, references, everything. You can still find archived items by filtering. Use archive for scripts or templates you don't need day-to-day but want to keep around.
Delete is permanent. It removes the resource and breaks its links to related objects. For example, deleting a script severs the connection from its submissions — those submissions still exist but lose their parent script reference.
Statuses
A submission moves through a few states. CREATED means it's queued and waiting to be dispatched to the cluster. SUBMITTED means Slurm has picked it up and is running (or scheduling) it. DONE means it completed successfully. REJECTED means Slurm couldn't accept the script — likely a syntax error or bad parameter. ABORTED means something went wrong during processing.
Submissions that stay queued longer than expected may indicate a busy cluster — check your cluster's utilization or pick a different partition.