Method and device for preventing cluster split-brain based on quorum server

Abstract

The invention discloses a high-availability cluster split-brain prevention method and device based on an arbitration server, which belongs to the high-availability cluster split-brain prevention technology in the technical field of computer clusters. In order to solve the problem that when the cluster heartbeat network is interrupted, the status of other nodes and their running services cannot be accurately judged, and the service cannot be taken over or the service is running on two nodes at the same time. The solution provided by the embodiment of the present invention includes: when the heartbeat network is interrupted, the cluster nodes that are not running the service can take over the service only after obtaining the corresponding service lock through the arbitration server, thereby avoiding the split-brain problem; after the service stops, the arbitration server reclaims the service lock and allow other cluster nodes to seize it again; in the process of multiple nodes simultaneously seizing the service lock, only one node successfully seizes and can start the service, preventing the occurrence of split-brain.

Description

technical field [0001] The invention belongs to the technical field of computer clusters, is applicable to high-availability clusters (High-availability Clusters), and particularly relates to the technical field of high-availability cluster split-brain prevention. Background technique [0002] With the rapid development of communication network technology, key fields such as telecommunications, finance, and e-government have higher and higher requirements for server availability. High availability (High Availability, HA) cluster technology can effectively reduce the service downtime of business systems caused by software and hardware failures. [0003] The current high-availability cluster system mainly uses links such as the network or serial lines as a private heartbeat network for communication between cluster nodes, which is responsible for exchanging information between synchronous nodes and monitoring the operation of each node in the cluster. When the service running...

AI Technical Summary

Problems solved by technology

However, the inventor found that these methods have limitations. In the actual environment, the hardware conditions for Fencing are often not available, and other important services are also running on the backup node. The customer does not allow the operating system to be restarted or the shared storage to be isolated.

In addition, although the disk lock technology based on the shared disk array can partially solve the cluster split-brain problem in the local area network and the occasion with the shared disk array, it also has many limitations, such as the need to re-partition the shared disk array, and does not support no disk array Environment, does not support virtual machine environment, does not support WAN remote cluster, etc.

Images