Prediction of chaotic systems relies on a floating fusion of sensor data (observations) with a numerical model to decide on a good system trajectory and to compensate non-linear feedback effects. Ensemble-based data assimilation (DA) is a major method for this concern depending on propagating an ensemble of perturbed model realizations. In this paper we develop an elastic, online, fault-tolerant and modular framework called Melissa- DA for large-scale ensemble-based DA. Melissa-DA allows elastic addition or removal of compute resources for state propagation at runtime. Dynamic load balancing based on list scheduling ensures efficient execution. Online processing of the data produced by ensemble members enables to avoid the I/O bottleneck of file-based approaches. Our implementation embeds the PDAF parallel DA engine, enabling the use of various DA methods. Melissa-DA can support extra ensemble-based DA methods by implementing the transformation of member background states into analysis states. Experiments confirm the excellent scalability of Melissa-DA, propagating 16,384 members for a regional hydrological critical zone assimilation relying on the ParFlow model on a domain with about 4M grid cells. The same use case was ported to the PDAF state-of-the-art DA framework relying on a MPI approach. A comparison with Melissa-DA at 2,500 members on 20,000 cores shows our approach is about 50% faster per assimilation cycle.