Kubernetes Resource Requests and Limits: How Misconfiguration Causes OOMKill Incidents

Kubernetes distinguishes between resource requests and limits, two settings that serve entirely different purposes and are enforced by different system components. Requests act as a scheduling promise, telling the Kubernetes scheduler how much CPU or memory to reserve on a node for a pod. Limits function as a hard runtime ceiling; if a pod exceeds its memory limit, the Linux kernel's OOM killer terminates it immediately without warning. Misconfigured or absent limits are among the most common causes of production outages, often resulting in pods entering a repeated crash-restart loop. Understanding how these settings interact with the scheduler and the container runtime is essential for preventing such failures in production environments.