Parallel drain in Cluster Autoscaler

[x13n]

Which component are you using?:

cluster-autoscaler

Is your feature request designed to solve a problem? If so describe the problem this feature should solve.:

Scale down of non-empty nodes taking very long time in large clusters.

Describe the solution you'd like.:

Parallel drain

Describe any alternative solutions you've considered.:

None really, the main factor contributing to slow scale down is the time to drain nodes one by one.

Additional context.:

Design was added in https://github.com/kubernetes/autoscaler/pull/4766

/assign