When Kubernetes Misbehaves – Pitfalls and Lessons Learned
summary
Live debugging recipe: add distributed tracing, metrics and logs to avoid blindfolded troubleshooting. Mizu tracing reveals the request hops and immediately shows the backend returning 500. Prometheus + Grafana surfaces that the backend is hitting 200% CPU of its request/limit so it’s simply overloaded. Loki logs then show “process time was too long,” confirming slow processing as the root cause. The demo installs these tools in-cluster and diagnoses the outage without touching application code.
AI-written summary · about
chapters (11)
- 0:00Welcome and introduction
- 1:10Why Kubernetes is complex
- 5:30Demo starts application issue
- 6:14Distributed tracing with Mizu
- 10:32Installing Prometheus metrics
- 13:44Key metrics to monitor
- 18:11Node failures and logs
- 21:58Importance of application logs
- 23:25Change tracking and root cause
- 24:50Building resilience over time
- 25:31Q&A session
related talks
01h 32mTutorial: Debug Your Kubernetes Apps
30mStreamlined Troubleshooting in Kubernetes
18mK8s Troubleshooting Demystified: Five Best Practices
50mTroubleshooting & Debugging Microservices in Kubernetes
36mKubernetes Confessions: Real Mistakes We’ve All Made (and What They Teach Us)
35mMonitoring Distributed Java Applications at Scale
30mMix Kubernetes with Linux Kernel Tracing to Boost the Observability of Your Cluster
14mLeveraging Distributed Tracing in Kubernetes & Containerized Environments