Fill Four On-Call Lists Before You Accept the Rotation

Описание и применение

Ротации теперь начинаются с полностью заполненного каталога, читаемых скриптов, четко определенного процесса эскалации и правила заморозки, что устраняет импровизацию и повышает безопасность.

Задача

On‑call rotations often start without fully defined runbooks, leading to improvisation, unclear escalation paths, and unsafe write permissions.

Результат

Rotations now have auditable, immutable runbooks that prevent accidental writes, clarify escalation paths, and reduce on‑call uncertainty.

Рабочий процесс

• 1. Verify every alert ID in the past month has a catalog entry, including a "not this" note.
• 2. Ensure each catalog row points to a read‑only first‑command script on disk.
• 3. Define escalation roles, minute bounds, and packet fields in escalation.yaml.
• 4. Set freeze_on_unknown_alert to true so unknown IDs trigger a freeze until a human edits the catalog.

Доказательства

I keep the on‑call file as four lists because each list answers a different failure.

Источник

Unknown IDs freeze writes immediately; unfreeze is a filled sentence, not a vibe.

Источник