HPA scale-in: is pod resource usage taken into account to prioritize pods to be deleted ?

[remiville]

Hello,

I know that in case of node resource pressure, the controller prioritizes pods to be evicted, for example in case of node memory pressure, pods using more memory than requested are prioritised for deletion.

But for HPA scale-in I struggle to find if and how pods are sorted to prioritize them for deletion.

While looking at the PodDeletionCost feature PR I found this priorization comment in the code https://github.com/ahg-g/kubernetes/blob/2d88d2d993b36c231fdc7537212685544469c517/pkg/controller/controller_utils.go#L785

Does it mean that no resource usage is taken into account when the controller has to select pods to be deleted during a scale in event ?

For example, if a scale-in occurs because the CPU usage is lower than the threshold set in the HPA, the controller will select pods to be deleted without prioritizing them by their current CPU usage, right ?

[k8s-ci-robot]

This issue is currently awaiting triage.

If a SIG or subproject determines this is a relevant issue, they will accept it by applying the triage/accepted label and provide further guidance.

The triage/accepted label can be added by org members by writing /triage accepted in a comment.

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes-sigs/prow repository.

[HirazawaUi]

/sig autoscaling

[ishaankalra]

Can I give it a try?

[HirazawaUi]

Can I give it a try?

Need to wait for confirmation and triage from the sig autoscaling team.