On-call runbook

The first hour of a Sev‑1, in four moves.

Same four moves whatever broke, so nobody improvises at three in the morning. Turn a step for the number we hold ourselves to.

T+0 · The page fires

Alertmanager pages the primary. No triage by committee, no waiting on a dashboard.

4 minutes, median ack

Unacknowledged after eight minutes, it escalates to the secondary and the duty lead together.

T+5 · Channel opens

One channel, one incident lead, one scribe. Everyone else stays out of it.

Opened automatically

The bot creates the channel, pins the roles and starts the timeline the moment the page is acknowledged.

T+15 · Status page

Customers hear it from us before they hear it from each other.

15 minutes, hard limit

We publish even when the cause is unknown. Saying we are looking counts as an update.

T+60 · Written update

Either it is resolved, or there is a paragraph explaining why it is not.

Every hour after that

The post-incident review is public within five working days, with the timeline unedited.

Nine Sev‑1s in the last twelve months. All nine reviews are published.