runbook
Instructions for the engineer on call, one entry per alert or common fault.
A runbook entry gives an alert's symptom, meaning and urgency, ordered checks, safe actions with how to verify them, when and to whom to escalate, and how to confirm the fix; each alert links to its entry.
Examples
The runbook entry for "node offline" walks the engineer from gh-b2's last health message and its dark status LED to a likely power fault.
The entry for "ingester rejecting messages" fires when the rejected count stays above zero for 5 minutes and lists which device IDs and versions to check.
See also