The transmission grid could be damaged by geomagnetic disturbances (GMDs), caused by a solar storm. The negative impacts of GMDs are possibly mitigated via a set of corrective actions (e.g., line switching, generator dispatch, and locating blocking devices). Making such decisions is challenging due to the uncertainty in the magnitude and orientation of GMDs and insufficient historical data. In this paper, we propose a two-stage distributionally robust optimization (DRO) model that determines the corrective actions such that the worst-case expectation of the total cost is minimized. For the solution approach, we propose a column-and-constraint generation (CCG) algorithm that solves mixed-integer second-order conic programs iteratively. Moreover, we derive a monolithic reformulation of the proposed DRO model under the triangle support set, which can be used to speed up the CCG algorithm. We test our solution approach to ‘epri-21’ and ‘uiuc-150’ systems.