随着人工智能在系统维护中的应用日益广泛，自动化事件响应工具展现出了显著的效能。这些工具能够自动检查告警、形成假设、查询遥测数据、关联最近部署并自行实施修复1。然而，这种高效的自动化解决方案正在带来一个隐忧：工程师正在逐步失去处理日常事故的实践机会，面对复杂和高危事故时的反应能力随之下降1。
早在1983年，研究者Lisanne Bainbridge就在论文《自动化的悖论》中指出了这一现象的本质——自动化减少了操作人员的日常练习机会，同时要求他们在异常情况下具备更强的能力1。为了应对这一挑战，业界开始借鉴其他高风险领域的做法。美国联邦航空局规定，商业航空公司机长必须每6个月完成一次反复训练或熟练度检查，包括起飞时发动机故障等应急场景1。
基于这些认识，一些技术团队已经开始采取行动。在Rootly与Uptime Labs的合作中，通过逼真的事件模拟训练来帮助工程师应对电商中断等真实场景1。此外，在Dropbox创建的课程项目中，通过让学生诊断和修复故障基础设施的方式被证明有效——动手学习相比被动指导能取得更好的学习成果1。这些实践表明，事件模拟训练可能是保持工程师应急响应能力的关键。
As artificial intelligence tools increasingly handle system incidents automatically, engineers risk losing the hands-on experience necessary to respond effectively when complex failures occur.1 AI-powered site reliability engineering systems can now autonomously check alerts, form hypotheses, query telemetry data, correlate recent deployments, and implement fixes without human intervention.1 However, this automation paradox—identified in Lisanne Bainbridge's 1983 paper "The Ironies of Automation"—creates a troubling dynamic: while reducing the frequency of routine operational tasks, it simultaneously demands that engineers possess stronger capabilities to handle exceptional circumstances.1
The aviation industry provides a cautionary model for addressing this challenge. Commercial airline pilots are required by the U.S. Federal Aviation Administration to complete recurrent training or proficiency checks every six months, including scenarios such as engine failures during takeoff.1 Drawing on similar principles, practitioners have begun implementing realistic incident simulation training to maintain engineering readiness. Rootly and Uptime Labs have partnered on training programs that expose engineers to realistic scenarios such as e-commerce outages.1 Additionally, hands-on learning approaches—such as the course model created at Dropbox where students diagnose and repair deliberately broken infrastructure—have proven more effective than passive instruction in developing practical troubleshooting skills.1
评论
还没有评论，欢迎留下第一条。