Failure prediction method and device for cluster system

Abstract

Embodiments of the present invention provide a failure prediction method and device for a cluster system. The method includes: performing cluster analysis on event sequence pattern sets according to failure event sets of the cluster system to obtain event sequence clusters; performing causal correlation analysis on event sequence clusters to obtain causal correlation path clusters; performing failure analysis on causal correlation path clusters to obtain The failure mode of the cluster system; according to the failure mode, the failure prediction of the cluster system is carried out. The failure prediction method and device of the cluster system provided by the embodiments of the present invention cluster the event sequence pattern set to obtain the causal correlation path cluster, thereby obtaining the failure mode of the cluster system, and predicting the failure of the cluster system according to the obtained failure mode. Through in-depth analysis and mining of cluster system logs, reliable failure rules for failure prediction can be obtained, thereby improving the recall rate of failure prediction.

Description

technical field [0001] Embodiments of the present invention relate to computer technology, and in particular to a failure prediction method and device for a cluster system. Background technique [0002] With the widespread application of cluster systems in large-scale productive computing platforms, failures and errors of cluster systems also occur frequently, and business interruption, data leakage or even data loss caused by cluster system failures or errors often bring heavy blows to users . Therefore, the importance of the security, reliability and availability of the cluster system has become increasingly prominent. [0003] In the existing technology, the failure events in the cluster system log are usually simply counted to obtain the failure event sequence association rules (referred to as failure rules), and then the failure prediction of the cluster system is performed according to the failure rules, thereby improving the security of the cluster system performanc...

AI Technical Summary

Problems solved by technology

[0004] However, because many redundant events are recorded in the cluster system log, such as failure events that occur without warning or failure events that occur infrequently, the reliability of the failure rules obtained based on this statistics is low.

Therefore, the prediction recall rate of the failure prediction method in the prior art needs to be improved.

Images