Key Takeaways What idle means: Idle Kubernetes resources are the gap between what your pods reserve and what they actually use, like a pod that requests 4 GiB of memory, uses 500 MiB, and still locks up the other 3.5 GiB you pay for. Why it hides: The scheduler and Cluster Autoscaler both work from requests, not real usage, which lets a cluster look full while its nodes sit mostly idle. How to find it: Compare requested CPU and memory against actual utilization over time, then flag workloads with no traffic and volumes no pod mounts.