运行服务器上的一种纠错技术研究和探讨

维护工作是软件项目中一个重要的部分,尤其是在永不停止的服务器上。如果在系统运行过程中有异常被抛出,维护人员需要去重现并分析这个异常。然而,服务器是不能停下来让维护人员去发现哪一次的业务请求使服务器出现了异常。所以,设计了一种在保持服务器工作状态的前提下,使用特定日志的模拟机制来帮助维护人员重现和找出问题所在的方法。这种方法很有效也很容易操作。将阐述模拟机制所用到的特定日志是如何被收集的,运用该模拟机制应做哪些重要工作及模拟机制的工作方式。所介绍的方法已经在作者参与开发的一个航空综合管理系统中成功应用。

Research and Exploration on Error-correction Technique in Non-stop Server

The maintenance work is an important part of a project,especially for a non-stop server.If there is an exception thrown,we need to reproduce and analyze it.However,the server cannot stop to let us find out which request causes the exception.So we design a simulation with the special logs to help us to reappear and catch the problem in development environment.This is more effective and easier to control.Also,the simulation can help us to analyze the system bottleneck and load reappearance.This paper explains how the special logs are collected,some important things we do for simulation and how simulation works as we expect.The simulation is already used as an important function in a 700 person month project,which will support one of the biggest airlines company of Japan which service can not stop for any second.