How does Kubernetes determine the deletion order of pods based on their QoS level?

Kubernetes uses the QoS (Quality of Service) level of pods to determine the order of their deletion in case of resource shortage. There are three QoS classes: Guaranteed, Burstable, and BestEffort. When deleting pods, those with the BestEffort class are deleted first, then Burstable, and finally Guaranteed. This is because Guaranteed pods have guaranteed resources and are more critical for application stability, while BestEffort pods are the least prioritized.

Example deletion order:

• BestEffort (least prioritized)
• Burstable
• Guaranteed (most prioritized)

Thus, Kubernetes aims to minimize impact on critical applications by deleting less important pods first.