Explore techniques for debugging Kubernetes services without disrupting their state in this 33-minute conference talk. Learn about nondestructive forensics methods for investigating application errors, resource leaks, and race conditions in distributed systems. Compare various debugging approaches, including detailed logging, debug sidecars, and Linkerd's "tap" feature. Discover how to gather crucial information without restarting pods, even when errors occur unexpectedly in production. Examine the potential of ephemeral containers and learn how to leverage direct access to nodes for detailed network debugging within a pod's network namespace. Gain valuable insights into maintaining the integrity of your Kubernetes environment while effectively troubleshooting complex issues.