GPU Info Schema
This document defines the schema for the GET /gpu-info endpoint, which returns detailed information about GPU resources and usage across the cluster.
GPUInfoResponse
Response for GET /gpu-info.
Properties
| Property | Type | Description |
|---|---|---|
| nodes | NodeInfo[] | List of nodes with GPU resources |
| pods | PodInfo[] | List of pods using GPU resources |
| prometheusMetrics | boolean | Whether Prometheus metrics are available |
NodeInfo Object
| Property | Type | Description |
|---|---|---|
| nodeName | string | Name of the Kubernetes node |
| hostname | string | Hostname of the node |
| totalVramMB | number | Total VRAM available on the node in MB |
| usedVramMB | number | Total VRAM used on the node in MB |
| gpus | GPU[] | List of GPUs on the node |
GPU Object
| Property | Type | Description |
|---|---|---|
| id | string | GPU ID (index) |
| uuid | string | Unique UUID of the GPU |
| vramTotalMB | number | Total VRAM of the GPU in MB |
| vramUsedMB | number | Used VRAM of the GPU in MB |
| gpuType | string | Type/Model of the GPU (e.g., "NVIDIA A100") |
PodInfo Object
| Property | Type | Description |
|---|---|---|
| id | string | Pod ID (UID) |
| podName | string | Name of the pod |
| nodeName | string | Name of the node the pod is running on |
| gpuUsage | GPUUsage[] | List of GPU usage details for the pod |
GPUUsage Object
| Property | Type | Description |
|---|---|---|
| id | string | UUID of the GPU being used |
| vramUsedMB | number | Amount of VRAM used by the pod on this GPU in MB |
Example
{
"nodes": [
{
"nodeName": "gke-cluster-pool-1-12345",
"hostname": "node-1",
"totalVramMB": 81920,
"usedVramMB": 40960,
"gpus": [
{
"id": "0",
"uuid": "GPU-1234-5678-90ab-cdef",
"vramTotalMB": 40960,
"vramUsedMB": 20480,
"gpuType": "NVIDIA A100"
},
{
"id": "1",
"uuid": "GPU-aaaa-bbbb-cccc-dddd",
"vramTotalMB": 40960,
"vramUsedMB": 20480,
"gpuType": "NVIDIA A100"
}
]
}
],
"pods": [
{
"id": "pod-xyz-789",
"podName": "model-service-deployment-abc",
"nodeName": "gke-cluster-pool-1-12345",
"gpuUsage": [
{
"id": "GPU-1234-5678-90ab-cdef",
"vramUsedMB": 10240
}
]
}
],
"prometheusMetrics": true
}