↳If you don't define resource constraints, a single memory-leaking Pod will consume the entire Worker Node's RAM, causing all other apps to crash.
■Requests (The Guarantee): What the Pod *needs* to boot. The Scheduler uses this to find a Node with enough capacity. If no Node has 256Mi free, the Pod stays Pending.
■Limits (The Hard Cap): What happens if the Pod exceeds its Limit?
- RAM: The Linux kernel instantly terminates the process with OOMKilled (Out of Memory).
- CPU: CPU is compressible. The kernel will not kill the pod; it will artificially throttle it. Your app will run incredibly slowly. *(Many engineers advise setting CPU Requests, but leaving CPU Limits empty to avoid unnecessary throttling).*
▲ LIVE INCIDENTYour Java application is repeatedly crashing and restarting every 5 minutes.