Method and device for detecting health status of database nodes

Abstract

The invention discloses a method and a device for detecting the health status of database nodes. Methods include: For any oracle? The RAC database node, in each sampling period, grabs the active session state of the database node in the first time period recently, and calculates the abnormal waiting session ratio of the database node in the sampling period according to the active session state; judges the abnormality Whether the proportion of waiting sessions is greater than the preset first threshold. If so, determine whether the database node has the most serious waiting session in the sampling period. A node that has been judged to be unhealthy for a predetermined number of sampling periods in succession, and if so, shut down and terminate the database node. The invention realizes the oracle? Automatic real-time detection of RAC database node health status.

Description

technical field [0001] The invention relates to the technical field of databases, in particular to a method and device for detecting the health status of database nodes. Background technique [0002] The introduction of the concept of Oracle Real Application Clusters (RAC, RealApplicationCluster) is to improve the high availability of the system, and achieve disaster recovery through cluster software and shared storage. Since each instance (instance) shares physical storage, but each has an independent memory structure, it is necessary to allocate some resources synchronously in real time. In some abnormal situations, it will cause poor node performance, and in severe cases, it will cause the database to hang (HANG). In the case of one of the nodes running out of performance and losing response, it will lead to the suspension of resource synchronization between nodes, affecting the normal operation of the entire database, and causing the business system to fail to provide s...

AI Technical Summary

Problems solved by technology

At this time, the database is generally in the stop responding or extremely slow response or HANG state, causing a large area of ​​business to be affected, and it is impossible to effectively and quickly discover performance problems, and then take further processing measures

[0010] Therefore, the original technical points are only simple and one-sided information collection and simple performance judgment, which are obviously not enough for complex OLTP large-scale database systems that require quick response.

Except for some specific and obvious system error messages, the evaluation of the health status of other database nodes largely depends on the experience and judgment of the maintenance personnel, and cannot be implemented in a standardized manner.

Images