In this paper, we investigate the dynamics of disaster spreading in community networks. Unlike the case in other kinds of complex networks, the transitions between the self-healing and spreading dynamics for community networks are in ladder-type in dependent of the recovery time and time delay. Statistical analysis on the average erupting time of disaster in communities shows that large-scale disaster firstly erupts in the community with maximal average degree, which is unrelated with the location where the initial disturbance seed is. Moreover, the average erupting time of disaster reduces along with the decrease of the average degrees of communities. Therefore, the community structure is helpful on the control of the disaster spreading.