This paper proposes a hierarchically coordinated control scheme for DC MG clusters under uncertainty. In each MG, the tertiary level controller optimizes the operating cost in the MG considering the real-time uncertainties of renewable generations and loads deviated from the forecasting data; and the primary controller responds to the real-time power fluctuations through an optimised droop curve. The hierarchically coordinated optimization problem is constructed to optimize the power set points and droop curve coefficients simultaneously under uncertainties using adjustable robust optimization model. For the MG cluster, the energy sharing of each MG in the cluster is optimized to minimize total operating cost and the transmission loss; the overall optimization model is solved in a distributed manner by alternating direction method of multipliers (ADMM) where each MG entity only exchange boundary information (i.e. the power exchange of MG entity with the MG cluster), thus information privacy and plug-and-play feature of each MG are guaranteed. The proposed approach optimally coordinates the operation and real-time control layers of a DC MG cluster with uncertainties; it achieves decentralized power sharing at real-time control layer and distributed optimization at operation layer, featuring high scalability, reliability and economy.