In VANETs (Vehicular Ad-hoc Networks), data can be transmitted to long-distance targets through multi-hop wireless broadcast, but the rapid changes in network topology and unstable wireless communication quality will cause performance unsteady for data transmission over VANETs. To meet the challenges, a reliable and connectivity-aware data dissemination mechanism which is applicable to urban scenarios is proposed. The proposed mechanism uses the traffic-flow data collected at intersections in realtime to build a Markov model of queue length change of vehicles, and then estimate the real-time varying of connectivity of wireless networks on the roads of two adjacent intersections so as to choose the minimum delay path to forward the data. For reliable data transmission, an improved greedy broadcast algorithm which merges into coverage elimination rule and multi-candidate strategy is also designed to ease the interference of urban environments. The simulation experimental results show that our mechanism can achieve high reliable and low-delay data dissemination in comparison with the epidemic-based protocol and the static-node-assisted adaptive data dissemination protocol.