Quick node replacement method for large-scale cluster database

Abstract

The invention provides a quick node replacement method for a large-scale cluster database. The method comprises the following steps of setting an unavailable state of a node; installing an installation program of a new node replacing the node, and performing new node and cluster configuration; setting a cluster mode to be read-only, and judging whether the execution of write operation is finished or not; after the execution of the write operation is finished, synchronizing metadata; and after the metadata synchronization is finished, recovering the cluster mode to be normal, setting a synchronous data flag for user data, and performing user data synchronization according to the synchronous data flag. According to the quick node replacement method for the large-scale cluster database, the data node replacement process is divided into two steps of metadata synchronization and user data synchronization; a user is prohibited to perform the write operation on the database during the metadata synchronization period, and the user is allowed to perform all database operations during the user data synchronization period; and the data volume of database metadata is not large in general, and the synchronization time is short, so that the time in which the user cannot perform the write operation on the database can be shortened to the maximum extent.

Description

technical field [0001] The invention relates to the field of database node replacement, in particular to a fast node replacement method for a large-scale cluster database. Background technique [0002] As the scale of the cluster database expands, the number of cluster nodes continues to increase, and damage to cluster nodes will become a common phenomenon. At the same time, as the amount of data increases, the computing power and storage capacity of a single node will also become a bottleneck. In both cases, it is necessary to replace and upgrade the cluster nodes so that the cluster can work normally, which is a typical application scenario of the present invention. [0003] Cluster data nodes can also be replaced by manual methods. The problem with this operation is: [0004] First, it is necessary to stop the current cluster and interrupt the user's business operations to wait for the administrator to perform the replacement operation. [0005] Second, when the amount...

AI Technical Summary

Problems solved by technology

[0004] 1. It is necessary to stop the current cluster and interrupt the user's business operations to wait for the administrator to perform the replacement operation

[0005] Second, when the amount of data is huge, it takes a lot of time to replace data nodes, and the user's business pause time will also be very long, which is unbearable for some users with high real-time requirements