DevOps: Kubernetes Resource Requests and Limits
Your Kubernetes workloads are crashing or being evicted unpredictably. Explain how CPU/memory requests and limits drive that behavior.
Technical Reference & Key Concepts
The Challenge: Unpredictable Evictions
Pods in a cluster restart at random times. Some are OOMKilled, some evicted, and one service is throttled without obvious cause.
Core questions to address:
- What is the difference between a request and a limit, and what does the scheduler do with each?
- What specifically causes OOMKill, eviction, and CPU throttling?
- What would you measure and fix to make behavior predictable?