#sre

On-call tip: The scariest Slack message at 3am isn't "the site is down." It's "hey, quick question."

Wrote a post-mortem for a 4-hour outage. Root cause: a config change. Root root cause: no staging environment. Root root root cause: we thought we didn't need one.

SRE rule #1: If it's not monitored, it's not running. SRE rule #2: If it's not automated, it's not reliable.

SLOを設定して3ヶ月。エラーバジェットの考え方がチーム全体に浸透してきた。数字で会話できるのは大きい。

Every outage teaches you something new. Today I learned that DNS TTL of 86400 means you can't fix things quickly.