DevOps Daily with Fexingo · 2026-08-15 · 12 min
In this episode of DevOps Daily, Lucas and Luna dig into a classic Kubernetes failure mode: kubelet eviction signals that silently kill pods. The trigger is often memory pressure on the node, and the fallout is worse than a crash - the pod vanishes without a proper termination, leaving no time for graceful shutdown. They walk through a real-world scenario from an e-commerce platform where the memory limit was set too high, the node ran out of memory, and the eviction manager picked victims in an order nobody expected. Lucas explains the difference between soft and hard eviction thresholds, and why the kubelet's eviction signal isn't just about memory - it can also be driven by disk pressure, inodes, or even PID limits. They also cover the critical role of the eviction manager's ranking by QoS class, and why Burstable pods often get hit first. The conversation wraps with practical advice: monitor eviction signals early, set realistic memory requests, and test your graceful shutdown path with real load. If you've ever lost a pod without a clear reason, this episode gives you the mental model to troubleshoot it fast.
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