Visual-Inertial Odometry (VIO) algorithms play a crucial role in navigation and localization applications. However, they still face challenges in dynamic and adverse weather environments. Existing datasets often fail to capture the characteristics of such environments, making it difficult to evaluate the robustness of current VIO methods. To address this issue, we propose three contributions in this paper. Firstly, we present a novel dataset that records the flight of a simulated unmanned aerial vehicle (UAV) in challenging dynamic and foggy conditions. The dataset includes five different scenes, each with distinct characteristics, where we introduce moving objects and vary lighting and weather conditions. It comprises synchronized images, IMU data, as well as ground truth trajectories. Secondly, we compare state-of-the-art VIO algorithms on our dataset, demonstrating significant performance degradation in the aforementioned scenes. Lastly, we summarize the limitations of current VIO algorithms and provide insights into future research directions.