This document discusses some of the challenges of running Apache Kafka at scale at LinkedIn, including issues with multitenancy, infrastructure, and management. It describes how high volumes of data and many producers can complicate ownership and capacity planning when data is shared. It also explains the pain points of tools like Mirror Maker and the lack of topic configuration management across clusters. Finally, it outlines some of LinkedIn's open source efforts to improve Kafka operations through tools like Cruise Control, Kafka Monitor, and kafka-tools.