To solve the rescheduling problem of discrete manufacturing shop with uncertain processing time, an improved ant algorithm based on the rolling horizon procedure is proposed. The dynamic scheduling strategy is based on a hybrid scrolling mechanism driven by events and cycles. Analysing disturbances in the process, we divide them into explicit and implicit disturbances. The length deviation tolerance (LDT) of processing time is designed and proposed to filter out redundant rescheduling. The dynamic scheduling algorithm, based on improved ant colony algorithms, is considered with the resource constraint of the shop, and with the use of a customized state transition rule, it helps to overcome the drawbacks of long ant path searching prone to stagnation. Using simulation, the performance of the dynamic scheduling strategy and scheduling algorithm are analysed and verified, and better rolling scheduling policy parameters are obtained.