DevOps Daily with Fexingo · 2026-08-01 · 9 min
In this episode, Lucas and Luna dig into a surprisingly common Kubernetes failure mode: when Pod PriorityClasses and preemption interact badly with cluster autoscaling, nodes get drained, workloads get killed, and the cluster autoscaler ends up fighting itself. They walk through a real-world scenario where a high-priority batch job preempts a low-priority web service, the node autoscaler scales down the now-idle node, and the preempted pods can't reschedule because the cluster is too small - creating a deadlock. Lucas explains how preemption is designed to work, why the autoscaler's node utilization threshold can trigger a scale-down right after preemption, and what signals you can look for in events and metrics to catch this before it bites. They also discuss practical mitigations: using PodDisruptionBudgets with preemption, tuning the autoscaler's scale-down utilization threshold, and considering whether preemption is even the right tool for your workload. If you run a cluster with mixed priorities and autoscaling, this episode gives you a concrete failure pattern to recognise and avoid.
How Kubernetes Namespace Quotas Trigger Unexpected Failures58 / 100Why Your Kubernetes Cost Reports Are Lying To YouHow Kubernetes Node Pools Can Cost You More Than You ThinkHow Kubernetes Pod Overhead Changes Node Capacity CalculationsHow Kubernetes Pod Security Admission Blocks Risky Workloads