This document discusses managing growth in production Hadoop deployments as more users and workloads are added to an initial Hadoop cluster. It begins by describing how an initial small cluster for ETL workloads can succeed but then struggle as data scientists, BI teams, and mobile teams also start using the cluster. The document then covers three main failure categories that can occur due to too much data, too many jobs, or too many users accessing the cluster. It provides examples of pressure points within these categories and recommends strategies to address them like using quotas, optimizing small files, tuning queues, and implementing access controls.