Edge computing seeks to enable applications with strict latency requirements by utilizing compute resources deployed closer to the users. The diverse, dynamic, and constrained nature of edge infrastructures necessitates a flexible orchestration framework that dynamically supports application QoS requirements. However, existing state-of-the-art orchestration platforms were designed for datacenter environments and make strict assumptions about underlying infrastructures that do not hold for edge computing. This work proposes a novel hierarchical orchestration framework specifically designed for supporting service operation over edge infrastructures. Through its novel federated cluster management, delegated task scheduling, and semantic overlay networking, our system can flexibly consolidate multiple infrastructure operators and absorb dynamic variations at the edge. We comprehensively evaluate our proof-of-concept implementation -Oakestra -against state-of-the-art solutions in both controlled and realistic testbeds and demonstrate the significant benefits of our approach as we achieve ≈ 10× and 30% reduction in CPU and memory consumption, respectively.