Kubernetes version: 1.19
Cloud being used: EKS
#CNI and version:
#CRI and version:
I am using “placeholder” pods with a negative priority in order to reserve space in the cluster for autoscaling (see Kubernetes Cluster Over-Provisioning: Proactive App Scaling - DEV Community for example)
With k8s 1.19, I now have the ability to define a set of “high but not critical” priority pods which will be scheduled with a higher priority than my normal pods, but won’t preempt any of the the normal pods (via PreemptionPolicy: Never). However, I do want these pods to preempt the “placeholder” pods that have negative priority.