The invention discloses a method and system for self-repairing down of a network server. The method comprises the following steps that: a, the server is started in a normal mode; b, the server starts to provide service, stores data to a shared memory, and simultaneously monitors whether the service is normal or not, if so, monitoring is carried out continuously, or otherwise, a step c is carried out; c, a restart script is called, and the server is started in a repair mode; d, the server uploads data from the shared memory and carries out data repair; and e, after the completion of repair, back to the step b. Due to the adoption of the method or system, a spare machine has no need to be configured for the server, so that the cost is low; when the server uploads data from the shared memory and carries out data repair, the repair speed is fast, the repair can be completed in a very short time, and further the service is recovered; and in addition, when the service needs to be migrated to other servers, reliable base guarantee for data replication can be provided, the data is transmitted to the other servers through reading the data of the shared memory, same service is started, so that working migration can be completed.