Failure Stories From the On-premise Bare-metal World

Explore real-world challenges and failures encountered when setting up large-scale Kubernetes clusters on-premise with bare-metal machines in this 31-minute conference talk. Dive into the specific network design considerations, compromises, and solutions implemented by Stephan Fudeus and David Meder-Marouelli from 1&1 Mail & Media Development & Technology GmbH. Learn about their experiences with clusters of up to 150 nodes, covering topics such as pod overlay networks, host network integration, traffic ingestion, and direction. Gain insights into the complexities of working with ingress controllers, BGP, network routing, DNS, iptables, and hardware load balancer appliances in a non-cloud-native environment. Understand the limitations and obstacles faced when implementing Kubernetes in on-premise, bare-metal setups and how to navigate these challenges.