Reconfiguring NoSQL databases in the face of changing workload patterns is crucial for maximizing database throughput. However, this is challenging because of the large configuration parameter search space with complex interdependencies among parameters. While stateof-the-art systems can automatically identify close-tooptimal configurations for static workloads, they suffer for dynamic workloads. This happens due to the following two fundamental limitations. First, they do not account for performance degradation due to database restarting (often needed to apply the new configurations), and second, they overlook the application's availability requirements during reconfiguration. Our solution, IRIS, addresses both these shortcomings and we demonstrate its effectiveness for a multi-tenant, global-scale metagenomics pipeline called MG-RAST and an HPC data analytics job queue, both of which have dynamically changing workloads.We compare the benefit of IRIS in throughput over the default, a static configuration, and a theoretically ideal solution.