Trial by Fire: Stress Testing Kubernetes Scalability Limits with kube-burner

Before you launch a massive workload to production, do you actually know if your Kubernetes cluster will survive?

Most engineering teams focus heavily on load-testing their application's HTTP endpoints, but they completely ignore the underlying platform. What happens to etcd, the API server, or your network plugin (CNI plugin) when hundreds of resources churn at once?

This talk introduces kube-burner, an open-source tool designed to aggressively stress-test your Kubernetes infrastructure by creating, updating, and deleting native resources at scale.

We are skipping the heavy theory and jumping straight into the terminal for practical, live demonstrations:

1. Demo 1: The Crash Test. We will flood a cluster with a high-density workload (deployments, secrets, configmaps) to see exactly when and how the control plane starts to struggle.
2. Demo 2: Live Observability. I will show you how to hook kube-burner into Prometheus to stream metrics like pod-ready latencies and etcd health directly into a live Grafana dashboard.
3. Demo 3: Deep-Dive Profiling with pprof. When the cluster chokes, why did it choke? We will use kube-burner to automatically collect pprof profiles (CPU and memory) from control plane , Cluster API server , components and even custom runtimes like CRI-O running on K8s node during the stress test to pinpoint exact code-level bottlenecks.
4. Demo 4: Bring Your Own Workload. A quick walkthrough on how to write declarative YAML configs to simulate and test your own specific production use cases (I would be demonstrating my custom workload to stress test a KEDA controller on RabbitMQ based auto-scaling application ).

1. Test the platform, not just the app: Learn how to stress-test your Kubernetes control plane (API server, etcd, CNI) to find its true breaking point.
2. Simulate real-world chaos: How to use simple YAML configurations to generate massive, high-churn infrastructure workloads.
3. Root-cause analysis with pprof: How to automatically extract CPU and memory profiles during a scale test to generate flame graphs and debug memory leaks in Go-based cluster components.
4. Visualize the bottlenecks: Practical techniques for scraping and graphing native cluster performance metrics (like exact pod startup latencies) to identify regressions.