SLURM Configuration Reference
The authoritative operating reference for your SLURM environment
Built for IT administrators and systems engineers operating a 3COL deployment, the Civic Compute SLURM Configuration Reference is a comprehensive document for your institution’s actual installation. It records the configuration decisions, resource model, and extension patterns your team needs to operate, maintain, and evolve the environment without relying on generic SLURM documentation.
What the reference covers
Your reference is organized around the SLURM controls and operating practices that matter in a licensed 3COL deployment.
- GRES (Generic Resource Scheduling): GPU resource definitions, counts, types, and binding conventions.
- QOS (Quality of Service): priority tiers, limits, and fairshare configuration that govern access to shared capacity.
- Accounting enforcement: CRAM integration, job billing, and the user and account structure required for controlled usage.
- Federation: multi-site SLURM federation setup for participation in the cooperative network.
- Provisioning automation: site-provisioning scripts and configuration templates for adding nodes, partitions, and users.
Partition reference
The document defines the access model and operating expectations for the four partitions used in 3COL deployments.
| Partition | Purpose | Notes |
|---|---|---|
| gpu-high | Priority jobs | Configured for time-sensitive or elevated-priority workloads. |
| gpu-normal | Standard jobs | Default shared capacity for routine instructional, research, and operational work. |
| gpu-college | College-reserved capacity | Capacity reserved for the licensed institution’s designated use cases. |
| gpu-community | Community and cooperative access | Introduced in Phase 2 for cooperative network participation. |
Engagement details
Fixed scope
80 hours
Deliverable
SLURM Configuration Reference document
Not-to-exceed
$6,000
Timeline
Coordinated with deployment or delivered as a standalone engagement