Abstract: In order to enhance the fault-tolerant and error-correcting ability of massive seismic streaming data acquisition and improve the stability and reliability of data processing, a distributed real-time seismic streaming data processing method based on ZooKeeper is proposed. ZooKeeper distributed service architecture was constructed, and the active and standby switching in the whole process was completed through the monitoring action, and the data consumption behavior was continued, which realized the functions of service registration, service loading, service fault tolerance and policy scheduling. The test results show that the average hourly service repair time (MTTR) of the system is reduced by 98.4%, the average hourly time between failures (MTTF) is increased by 33.5%, and the data integrity rate is increased by 20% compared with the failure scenario, which can better solve the stability and reliability of seismic streaming data. It has the potential of popularization and application of regional seismic network technology.