cluster

type ClusterInfo struct {
	Platform        string `json:"platform"`
	GPUArchitecture string `json:"gpuArchitecture"`
	GPUProduct      string `json:"gpuProduct"`
	// GpusPerNode is the smallest GPU count any discovered node reports. A job
	// asks for the same count on every node, so the smallest node sets what
	// the cluster can run.
	GpusPerNode int32 `json:"gpusPerNode"`
	// CatalogGpusPerNode is what the catalog assumes for this architecture. It
	// can differ from GpusPerNode, for example on a workstation or a partly
	// drained node, and then a certification must set gpusPerNode itself.
	CatalogGpusPerNode int32         `json:"catalogGpusPerNode"`
	TotalNodes         int           `json:"totalNodes"`
	TotalGPUs          int           `json:"totalGPUs"`
	Nodes              []NodeInfo    `json:"nodes"`
	Topology           *TopologyInfo `json:"topology,omitempty"`
}

type NodeInfo struct {
	Name string `json:"name"`
	// GPUs is the node's allocatable nvidia.com/gpu count. It is 0 when the
	// device plugin has not advertised the resource yet.
	GPUs  int32  `json:"gpus"`
	Ready bool   `json:"ready"`
	Rack  string `json:"rack,omitempty"`
}

type RackInfo struct {
	Name      string   `json:"name"`
	Nodes     []string `json:"nodes"`
	NodeCount int      `json:"nodeCount"`
}

type TopologyInfo struct {
	Key   string     `json:"key"`
	Racks []RackInfo `json:"racks"`
}