Dashboard panel for node's available ephemeral storage space

[consideRatio]

A node can evict pods based on memory pressure, but also on using up too much ephemeral storage space which is any space outside mounted folders. This could for example be /tmp but perhaps not /home/jovyan.

Typically pods doesn't request ephemeral storage space explicitly, but they can do this also just like they can request it for CPU and Memory, and a node has a capacity and allocatable amount of ephemeral storage space just like they have for CPU and Memory.

Since too little ephemeral storage left on the node can trigger pod evictions, and is relevant for deciding how large node disks should be used when setting up node pools, it would be great to have a dashboard to inform us ephemeral storage use just like we have for a node's ram usage.

Related

GCP Nodes doesn't make all the disk space available for pods as ephemeral storage, only ~50% of it. Consider for example n2-highmem-16 and -64 nodes without explicitly sized disk nodes, that has defaulted to a 100GB disk size.

  n2-highmem-16:
    allocatable:
      cpu: 15890m         
      ephemeral-storage: "47060071478"
      memory: 122210676Ki
    capacity:
      cpu: "16"
      ephemeral-storage: 98831908Ki
      memory: 131919220Ki
  n2-highmem-64:
    allocatable:
      cpu: 63770m
      ephemeral-storage: "47060071478"
      memory: 510603916Ki
    capacity:
      cpu: "64"
      ephemeral-storage: 98831908Ki
      memory: 528365196Ki