Kubernetes scheduling deep dive: Priority, preemption, and resource requests

When a high-priority pod can’t schedule, Kubernetes doesn’t just wait – it actively evicts lower-priority pods to make room. Understanding when preemption happens, which pods get evicted, and how resource requests influence these decisions is essential for running stable multi-tenant clusters.

This session explores the intricate relationship between pod priority, preemption logic, and resource requests in the Kubernetes scheduler.

We’ll trace what happens when a pod fails to find a feasible node: how the scheduler evaluates every node for potential victims, how it respects Pod Disruption Budgets, and why your resource requests determine whether you’re safe from preemption.

You’ll learn how QoS classes interact with priority-based preemption and why setting requests is your first line of defense against unexpected evictions.

Topics covered include:

• Understanding the scheduler’s preemption process
• How resource requests and QoS influence scheduling decisions
• Protecting workloads with priority classes and Pod Disruption Budgets
• Troubleshooting common preemption and eviction scenarios

Who is this for?

Platform engineers, SREs, and cluster administrators who need to understand how Kubernetes makes

scheduling decisions under resource pressure.

Who is the speaker?

Daniele is an instructor at LearnKube, teaching Kubernetes and containers to small and large enterprises.