Discussion: InstasliceSpec for MemoryOnNodeAtBoot and MemoryOnCpuAtBoot

[kannon92]

// InstasliceSpec defines the desired state of Instaslice
type InstasliceSpec struct {
	// migGpuUuid represents uuid of the mig device created on the gpu
	// +required
	MigGPUUUID map[string]string `json:"migGpuUuid"`
	// allocations represents allocation details of user workloads
	// +optional
	Allocations map[string]AllocationDetails `json:"allocations"`
	// migplacement represents gpu instance, compute instance with placement for a profile
	// +required
	Migplacement []Mig `json:"migplacement"`
	// cpuonnodeatboot represents total amount of cpu present on the node
	// +required
	CpuOnNodeAtBoot int64 `json:"cpuonnodeatboot"`
	// memoryonnodeatboot represents total amount of memory present on the node
	// +required
	MemoryOnNodeAtBoot int64 `json:"memoryonnodeatboot"`
}

We should discuss ways that Instaslice "knows" the amount of cpu and memory on node.

There is work being done in Kubernetes to allow for dynamic resizing of memory/cpu at the node level.

kubernetes/enhancements#3953

I also question why this should be a user facing API.