A hands-on tutorial using KubeLab, an open-source Kubernetes failure simulation lab, to explore seven real failure scenarios on a 3-node cluster. Covers pod deletion and ReplicaSet self-healing, node draining with PodDisruptionBudgets, CPU throttling via CFS limits, OOMKill behavior and silent restart loops, StatefulSet database failure with PVC persistence, cascading pod failure, and readiness probe failures. Each simulation includes kubectl commands to observe the failure, explains the underlying Kubernetes mechanics, highlights common production traps, and provides concrete fixes. Also covers key Grafana panels and Prometheus queries for detecting silent OOMKills and CPU throttling in production.
Table of contents
Table of ContentsWhat is KubeLab?PrerequisitesHow to Get the Lab RunningSimulation 1: Kill Random PodSimulation 2: Drain a Worker NodeSimulation 3: CPU Stress and ThrottlingSimulation 4: Memory Stress and OOMKillSimulation 5: Database FailureSimulation 6: Cascading Pod FailureSimulation 7: Readiness Probe Failure4. How to Read the Signals in Grafana6. How to Use This for Production DebuggingConclusionSort: