Priority-based scheduling assigns each Pod an integer priority through a PriorityClass. Kubernetes uses that value to rank Pods for scheduling, eviction, and, when configured, preemption. Priority helps critical workloads compete for limited cluster capacity, but it does not protect a Pod that exceeds its resource requests. section 5.1.3
When resources are constrained, Kubernetes applies these rules:
- Best-effort Pods are especially vulnerable because they have no resource requests. section 5.1.3
- Among Pods using more than requested resources, priority is considered before excess usage. section 5.1.3
Assigning priority to a Deployment
Create a PriorityClass such as high-priority with value 1000000, then set priorityClassName: high-priority in the Deployment’s Pod template. Use preemptionPolicy: Never when the priority should rank workloads without displacing lower-priority Pods.
