Distributed system multi-node locking method

Abstract

The invention discloses a multi-node locking method for a distributed system. The multi-node locking method comprises the following steps: 1, allocating identity information to all nodes to be locked;2, sending a pre-locking request to all nodes to be locked; 3, judging whether all the to-be-locked nodes are successfully pre-locked or not, and when the to-be-locked nodes fail to be pre-locked, sending a pre-locking release message, pre-locking failure and waiting time T1 to each to-be-locked node which is successfully pre-locked, and returning to the step 2, when all the to-be-locked nodes are successfully pre-locked, sending a locking request to all the to-be-locked nodes, and if the pre-locking is successful, executing the step 4; and 4, judging whether all the to-be-locked nodes are successfully locked or not, and when the to-be-locked nodes fail to be locked, sending an unlocking message, locking failure and waiting time T2 to each successfully locked to-be-locked node, and returning to the step 2. When all the to-be-locked nodes are successfully locked, the locking is successful. The method is not restrained by authoritative nodes, and the problems of single-point failure anddeadlock do not exist.

Description

technical field [0001] The invention belongs to the technical field of distributed systems, and in particular relates to a multi-node locking method of a distributed system. Background technique [0002] During the use of a distributed system, some operations need to lock resources to ensure that they are not interfered by other operations, thus affecting the running results. Different from the stand-alone system, each operation can determine whether the resource lock request is made earlier by the system time, so as to determine that the operation can lock the resource first. In a distributed system, resource locking operations are propagated in the network in the form of locking request messages. There is no standard time for reference among distributed nodes, so it is not easy to determine which operation sends resource locking requests first. A common locking mechanism in a distributed system is that an authoritative node manages the locks of all resources. The operatio...

AI Technical Summary

Problems solved by technology

However, there is a single-point bottleneck in this centralized management method. Even if the distributed lock mechanism such as Zookeeper solves the single-point bottleneck problem to a certain extent, during the master selection period of this mechanism, there is no authoritative node, and it is impossible to realize the The locking of resources not only increases the maintenance cost of the authoritative node, but also cannot use the distributed lock service during the switching process of the authoritative node

Images