solo engineer 没有 incident response,我自己就是 responder
团队有 incident response。
Incident response:事故响应。On-call:每周 7×24 轮值。Pager:电话响。Sev1 / Sev2 / Sev3:严重等级。War room:Google Meet 拉一群人。Blameless postmortem:不追责的事故复盘。Incident commander:协调事故。SLO / error budget:错误预算。
每个 startup 1 个 on-call rotation。PagerDuty $20/月每人。
solo engineer 没有 incident response。
我自己就是 responder。
我的散文挂了 = incident。我自己立刻知道(Vercel 监控)。我自己立刻 revert。不需要 pager、不需要 war room、不需要 blameless postmortem、不需要 incident commander。
我没 on-call 的必要——我不轮值。我没 pager 的必要——我自己听到。我没 Sev1 / Sev2 / Sev3 的必要——只有 0(挂了)或 1(好的)。我没 war room 的必要——我一个人,不需要房间。我没 blameless postmortem 的必要——commit log 是 postmortem。我没 incident commander 的必要——我 = commander。我没 SLO / error budget 的必要——我没承诺 SLO。
团队的 incident response 价值是「快速恢复 + 减少损失 + 学习」。前提是 7×24 SLA、需要快速恢复、需要合规。solo engineer 没这个问题——我没 7×24 SLA、我立刻恢复、commit log = learning。
一人工程的 incident response = git revert。不是 PagerDuty $20/月。