Distributed storage system, data recovery method and device thereof, and medium

Abstract

The invention discloses a distributed storage system, a data recovery method and a device thereof, and a computer readable storage medium. A plurality of storage nodes used for carrying out redundantstorage on the same data in the distributed storage system serve as a redundant backup group. The method is applied to a main storage node in a redundant backup group, and comprises the following steps: receiving a data acquisition request sent by a slave storage node in the redundant backup group; reading each piece of data needing to be sent to the slave storage node from the local; judging whether each piece of data is abnormal so as to determine the abnormal data as target data and postpone the sending of the target data; judging whether a target slave storage node storing complete backupof the target data exists in the redundant backup group; and if the target slave storage node exists, performing local data recovery based on the complete backup of the target data, and sending the locally recovered target data to the slave storage node. According to the method, a data recovery mechanism can be reasonably optimized, server hang-up is reduced, and the stability and sustainability of the service are improved.

Description

technical field [0001] The present application relates to the technical field of distributed storage, and in particular to a distributed storage system and its data recovery method, device and computer-readable storage medium. Background technique [0002] A distributed storage system is a storage system that divides data according to certain rules and stores them in multiple independent general-purpose storage servers. [0003] In order to ensure the security of user data, multiple copies of the same data are usually copied and stored on storage devices of different storage nodes, and these storage nodes with the same data backed up constitute a redundant backup group. For example, if the three-copy backup rule is adopted, one data is copied into three and stored in three different hard disks, and the redundant backup group includes three hard disks. Data maintenance can be realized by deploying PlacementGroup (PG) service on each storage node of each redundant backup grou...

AI Technical Summary

Problems solved by technology

However, in the data recovery mechanism in the prior art, the main storage node does not judge the abnormal data after reading the data from the local, but directly sends out

Once the storage node receiving the data receives the abnormal data, it will think that it has lost the data consistency, and then hang up the server of the storage node, resulting in the interruption of storage services, which intensifies the work pressure of other storage nodes in the distributed storage system , especially when the storage system is busy

Moreover, in general, a distributed system will automatically re-initiate a round of master node election after a storage node hangs up, which further affects the work efficiency of the distributed storage system

Images