KubeCon + CloudNativeCon India 2025

Are you experiencing performance issues with your Kubernetes control plane? Symptoms such as rising CPU/memory usage, 429 TooManyRequests errors, and delayed responses from the APIServer or ETCD can significantly degrade cluster responsiveness, especially in large environments, ultimately threatening infrastructure reliability.
In this talk, based on our experience managing tens of thousands of production clusters, we will explore how to leverage observability to identify performance bottlenecks and root causes. We'll compare logging, monitoring, and tracing within the Kubernetes control plane, using real-world example - analyzing LIST request.
We will introduce best practices for observability, discuss common performance pitfalls such as APF flow control issues, misconfigured APIServer webhooks, and the heavy burden of excessive LIST requests on control plane, etc. Additionally, we will share optimization solutions for these scenarios, empowering you to enhance cluster performance.