This contribution details the deployment of Rucio, the ATLAS Distributed Data Management system. The main complication is that Rucio interacts with a wide variety of external services, and connects globally distributed data centres under different technological and administrative control, at an unprecedented data volume. It is therefore not possible to create a duplicate instance of Rucio for testing or integration. Every software upgrade or configuration change is thus potentially disruptive and requires fail-safe software and automatic error recovery. Rucio uses a three-layer scaling and mitigation strategy based on quasi-realtime monitoring. This strategy mainly employs independent stateless services, automatic failover, and service migration. The technologies used for deployment and mitigation include OpenStack, Puppet, Graphite, HAProxy and Apache. In this contribution, the interplay between these components, their deployment, software mitigation, and the monitoring strategy are discussed.