GPU Inventory

Accessing the GPU Inventory Page

Project mode: Scheduler > GPU Inventory

1. Summary Cards

Card	Meaning
Node	Number of GPU nodes
Total GPU	Total physical GPU chips
behavior	Number of allocated GPUs
Idle occupancy	Allocated but unused GPUs
Unallocated	Free/unallocated GPUs
Operating rate	Percentage of GPUs in operation with a green progress bar

2. Filter Bar

Filter	Description
CLUSTER	Cluster dropdown, defaults to "entire" (all)
GPU MODEL	GPU model dropdown, defaults to "entire"
GPU STATUS	GPU status dropdown, defaults to "entire"
TYPE	GPU sharing-mode dropdown, defaults to "entire"
Search box	Searches "Node · UUID · Workload Search"

3. Explorer Tab - Left-hand tree

• "Tree filter…" input above the tree.
• Hierarchy: Cluster → Node → GPU, e.g. skpc-prod-cluster (11N) → Node-192-168-56-217 (x8) → GPU 0 … GPU 7.
• Each GPU row shows its model (e.g. "H100 · NVLink")
• Status indicators: | Color | Meaning | | --- | --- | | 🟢 Green | Healthy / Operating normally | | 🟡 Yellow | Warning condition | | 🔴 Red | Critical issue detected | | ⚪ Gray | Idle, disconnected, or status unavailable |
• Selecting a GPU in the tree highlights that row (light purple background) and shows its detail on the right.

3.1. Cluster Detail

• Cluster detail information:

Field	Description
Cluster Name	Name of the selected cluster.
Node	Total number of nodes in the cluster.
GPU	Total number of GPUs in the cluster.
In Operation	Number of GPUs currently running workloads.
Idle	Number of GPUs currently not assigned to workloads.

• Node List Table

Column	Description
Node	Node hostname or identifier.
GPU	Total GPUs available on the node.
Average Operating Rate	Average GPU utilization percentage across all GPUs on the node.

3.2. Node Detail

• Node Detail Information

Field	Description
Node Name	Name of the selected node.
Cluster	Cluster to which the node belongs.
GPU	Total number of GPUs installed on the node.
In Operation	Number of GPUs currently running workloads.
Idle	Number of GPUs currently not assigned to workloads.
Error	Number of GPUs currently reporting errors or abnormal conditions.

• GPU List Table

Column	Description
GPU	GPU identifier within the node.
Mode	GPU operating mode or allocation mode.
Situation	Current operational status of the GPU.
Operating Rate	Current GPU utilization percentage.
Temperature	Current GPU temperature in degrees Celsius.

3.3. GPU Detail

• Outline Tab:

Field	Description
GPU ID	Identifier of the selected GPU.
Model	GPU model and interconnect type.
Status	Current GPU operational status.
Type	GPU allocation mode.
Location	Physical location of the GPU within the cluster hierarchy.

• Characteristic Tab:

Field	Description
Operating Rate (SM)	Current streaming multiprocessor utilization.
Memory	Current GPU memory utilization.
Temperature	Current GPU temperature.
Electrical Energy	Current GPU power consumption.

• Workload Tab:

Column	Description
Workload	Name of the workload currently assigned to the GPU.
Project	Project associated with the workload.
Grass	Resource pool, namespace, or execution environment.
Situation	Current workload status.

4. Topology Tab

The topology view provides a visual representation of all nodes and GPUs within the inventory.

Each node card contains:

Field	Description
Node Name	Hostname of the node.
GPU Count	Number of GPUs installed on the node.
GPU Blocks	Visual indicators representing individual GPUs.
Status Indicator	Color-coded status of each GPU. • Green: In Operation (GPU is actively running workloads.) • Orange: Idle Occupancy (GPU is allocated but currently idle.) • Gray: Unallocated (GPU is available and not assigned.) • Red: Uncle (GPU has reached the maximization.)

The GPU Detail panel appears when a user clicks an individual GPU block within the topology view.

Field	Description
Model	GPU model and interconnect type.
GPU Identifier	Unique GPU identifier.
Allocation Type	GPU allocation mode.
Status	Current operating status.
Operating	Current GPU utilization
Memory	Current memory consumption

Detailed Information

Field	Description
Node	Host node where the GPU is installed.
Index	GPU index within the node.
GPU Pool	GPU resource pool assigned to the workload.
Namespace	Kubernetes namespace or project namespace.
Workload	Current workload using the GPU.

5. Inventory Tab

Column	Description
Node	Host node containing the GPU.
Model	GPU model and interconnect type.
Type	GPU allocation type.
UUID / Index	GPU index or unique identifier within the node.
Situation	Current operating status of the GPU. • Running: GPU is actively executing a workload. • Idle: GPU is allocated but not currently being utilized. • UNCLE: GPU is in an abnormal state and requires investigation. • Error: GPU has reported a failure condition.
Operating Rate	Current GPU utilization percentage.
Grass	GPU pool or resource group assigned to the workload.
Workload	Workload currently consuming the GPU resource.

The GPU Detail side panel is displayed when a user clicks a GPU row from the Inventory table.

Field	Description
Model	GPU model and interconnect type.
GPU Identifier	Unique GPU identifier.
Allocation Type	GPU allocation mode.
Status	Current GPU state.
Operating Rate	Current GPU utilization percentage.
Memory	Current memory consumption versus total capacity.

Detailed Information

Field	Description
Node	Host node where the GPU is installed.
Index	GPU device index within the node.
Temperature	Current GPU temperature.
Electrical Energy	Current power consumption.
GPU Pool	Resource pool assigned to the GPU workload.
Namespace	Namespace or project using the GPU.
Workload	Workload currently running on the GPU.