Slurm Plugin

The Slurm plugin integrates HPC workload manager data with physical infrastructure views. It is enabled by setting plugins.slurm.enabled: true in config/app.yaml. Full configuration lives in config/plugins/slurm/config.yml.

Plugin ID: workload-slurm

API Routes

Method	Path	Description
GET	/api/slurm/rooms/{room_id}/nodes	Node states mapped to room layout
GET	/api/slurm/summary	Cluster-wide Slurm summary
GET	/api/slurm/partitions	Per-partition statistics
GET	/api/slurm/nodes	Flat node list with topology context
GET	/api/slurm/mapping	Read current node name mappings
POST	/api/slurm/mapping	Save node name mappings
GET	/api/slurm/metrics/catalog	List loaded Slurm metric definitions
POST	/api/slurm/metrics/catalog/config	Update which metric files to load
GET	/api/slurm/metrics/data?metric_id=X	Query Prometheus for a Slurm metric

Menu Section

Contributes a Workload section (order=50) to the sidebar:

• Overview (/slurm/overview)
• Nodes (/slurm/nodes)
• Alerts (/slurm/alerts)
• Partitions (/slurm/partitions)
• Wallboard (/slurm/wallboard) — multi-room view with selectable display modes and Configure panel

Configuration

Full configuration lives in config/plugins/slurm/config.yml (config/app.yaml only controls plugins.slurm.enabled):

# config/plugins/slurm/config.yml
metric: slurm_node_status
label_node: node
label_status: status
label_partition: partition

roles: [compute, visu]
include_unlabeled: false
mapping_path: config/plugins/slurm/node_mapping.yaml

status_map:
  ok:   [idle, allocated, alloc, completing, comp, mixed, mix]
  warn: [maint, planned, reserved, drain, power_down, power_up, reboot_issued]
  crit: [down, fail, error, unknown, noresp, inval]
  info: []

severity_colors:
  ok: '#22c55e'
  warn: '#f59e0b'
  crit: '#ef4444'
  info: '#3b82f6'

metrics_catalog_dir: config/plugins/slurm/metrics
metrics_catalogs: [metrics.yaml]

Node Mapping

Node name mapping supports wildcards — no need to list every node individually:

# config/plugins/slurm/node_mapping.yaml
mappings:
  # Pattern: n001 → compute001, n002 → compute002, etc.
  - node: "n*"
    instance: "compute*"
  # Exact override for edge cases
  - node: "login01"
    instance: "service001"

The * wildcard matches any suffix; exact entries have higher priority. Mappings can also be managed from the UI: Settings → Plugins → Slurm → Edit mappings.

Metrics Catalog

Add YAML files under config/plugins/slurm/metrics/ to expose your exporter metrics in dashboards and tooltips. No code changes needed — register new files via metrics_catalogs.

# config/plugins/slurm/metrics/metrics.yaml
metrics:
  - id: slurm_running_jobs
    name: Running Jobs
    metric: slurm_running_jobs_total
    scope: global
    display: { unit: jobs, chart_type: gauge }

  - id: slurm_node_cpus_alloc
    name: Node CPU Allocated
    metric: slurm_node_cpu_alloc
    scope: node
    display: { unit: cores, thresholds: { warn: 80, crit: 95 } }

Device Role Filtering

Templates declare a role field to filter devices shown in Slurm views:

templates:
  - id: compute-node
    type: server
    role: compute   # compute | visu | login | io | storage

Dashboard Widgets

When the Slurm plugin is enabled, three widgets become available in the Dashboard Widget Library:

Widget type	Group	Description
slurm-cluster	Overview	Node state bar + severity breakdown
slurm-nodes	Stats	Total Slurm node count
slurm-utilization	Charts	Allocated % gauge

These widgets live in frontend/src/app/plugins/slurm/widgets/ and are hidden automatically when the plugin is disabled. They use requiresPlugin: 'slurm' — see the Dashboard Widget System guide.