What They Don't Tell You in the Docs: Surviving Day-1 Configs and Day-2 Outages

Production Kubernetes breaks in ways the documentation never warns you about. This talk reveals the real-world troubleshooting knowledge gained from helping customers survive Day-2 outages. Broadcom solution architects Pratik Jan and Saurabh Tandon share hard-won lessons from production incidents: MTU misconfigurations that silently kill large TCP transactions while small requests succeed, DNS lookup failures that appear intermittent, and image pulls that mysteriously fail after node refreshes. Learn how Day-0 and Day-1 configuration choices—network policies, multi-zone deployments, security postures—create hidden failure modes that only surface at 2 AM when your cluster is down. The talk focuses on MTU as a "silent killer" in Kubernetes networking, explaining why overlay networks (Geneve, VXLAN) cause packet drops that leave no trace, and how to spot the symptoms when autoscalers provision new nodes with mismatched MTU settings. Recorded at a Cloud Native event in Kerala, this session fills the gaps between "kubectl apply" and real production operations. Links: #Kubernetes #ProductionKubernetes #Day2Operations #NetworkTroubleshooting #MTU