Control method for cluster timing scheduling tasks

Abstract

The invention discloses a control method for cluster timing scheduling tasks. According to the control method, the time intervals of executing the tasks are controlled within a certain scope in the task scheduling process, a database only needs to maintain information like scheduling time and locking states of each timing scheduling task, and the timing scheduling tasks are executed without the limitation of nodes. According to the control method, the problem that the timing scheduling tasks are executed repeatedly due to desynchrony of machine time can be avoided, and the control method brings advantages of high concurrency and high stability of a cluster system into full play and can greatly reduce application embeddedness.

Description

Technical field [0001] The present invention relates to task scheduling technology of a cluster system, in particular to a method for controlling cluster timing scheduling tasks. Background technique [0002] With the development of computer technology, more and more services and applications need to be deployed in server clusters, which means that an application will be deployed on multiple machines or a machine will deploy multiple instances of the same application (also known as node). On the other hand, regular scheduling processing is often required in applications, such as bank interest settlement at 1:00 am every day, or e-commerce system statistics on merchandise sales every few hours. [0003] The combination of the above two scenarios results in such a problem: usually scheduled tasks can only be executed once at a specified scheduling time (for example, interest settlement scheduling tasks need to be settled once at 1 am), but if each application node does not The task...

AI Technical Summary

Problems solved by technology

[0003] The combination of the above two scenarios leads to such a problem: Usually, the scheduled scheduling task can only be executed once at a specified scheduling time (for example, the interest settlement scheduling task needs to perform an interest settlement at 1 am), but if each application node does not The task locking mechanism will trigger the execution of scheduled tasks at the specified time point

[0010] 1. Quartz's cluster scheduling control needs to create 12 tables in the database, which is relatively large for application embedding. Although the function is powerful, the configuration is also quite complicated, and the data is stored in binary, so it cannot be directly read and understood to check the running status

[0011] 2. If the scheduled task can be executed in a very short time, the time of each machine in the cluster is not synchronized, which will lead to the problem that the scheduled task is executed multiple times

It can be seen that because the time of node 1 and node 2 is inconsistent, and the execution time of the scheduled task is very short (only 30 seconds), when node 1 acquires the task lock, node 2 does not compete to acquire the lock at the same time to achieve mutual exclusion

After node 1 completes the task processing and releases the lock, node 2 also acquires the task lock, resulting in the second execution of the scheduled task

[0012] It can be seen that the existing control methods for cluster scheduled scheduling tasks have problems such as the inability to take advantage of the high stability of the cluster system or the high degree of application embedding and the multiple execution of scheduled scheduling tasks.

Images