Google Rewrites the SRE Book for AI-Era Reliability Engineering
Google’s second edition of its SRE book updates a first edition that standardized practices such as SLOs, error budgets, toil reduction and blameless postmortems. Authors told InfoQ 中文 AI&大模型 that platform engineering and AI are changing reliability work, including managing nondeterministic model outputs and reviewing AI-generated incident responses, so the rewrite emphasizes first principles, organizational culture and practices from other companies. The book also gives greater space to social-technical systems and the business value of reliability while trimming technical details that have become commodity features or easier to generate with LLM tools.
The rewrite offers SRE practitioners a framework for production AI systems, where conventional web-service assumptions such as fully measurable and predictable behavior may not hold.

编译|林绮蓓
策划|Tina
2016年,一本蓝色封面的书砸进了整个技术行业,把“运维”这两个字从地下室拽进了工程学科的殿堂。

在它之前,运维在大多数组织里是一个成本中心,是一群被凌晨告警叫醒的人,是开发团队眼中“拖慢发布速度的那帮人”,是一门靠师傅带徒弟、靠口口相传、靠个人英雄主义在机房里临场发挥的手艺。系统出了故障,运维的默认答案就是:“加人、加班、加流程”。
#Google#Site Reliability Engineering, Second Edition#Gemini CLI#Komodor#Spotify#Microsoft#Etsy#Disney#Waymo