On-call escalation policy
The rules for paging the primary responder, waiting for an acknowledgement, then paging the next person if nobody answers.
See it
What it is
An on-call escalation policy is the ladder a page climbs when nobody acknowledges it. Level one might notify the primary responder; after a fixed delay it pages the secondary, then a team lead or incident commander. The rotation decides who is on call right now. The escalation policy decides what happens after the alert starts.
Use one for every service important enough to wake a person. A useful policy names a staffed schedule at each level, sets a short but realistic acknowledgement timeout, defines when the ladder repeats, and gives responders a runbook. PagerDuty models this directly, with schedules, overrides, acknowledgement timeouts, and numbered escalation levels.
Gotcha: a beautiful ladder still fails if every level points to the same person, an empty schedule, or a phone left in do-not-disturb mode. Test the path with a real page, cover vacations and time zones, and keep contact methods current. Acknowledgement should stop duplicate paging, but it does not mean the incident is resolved, so track resolution separately.
Ask AI for it
Create PagerDuty on-call routing as Terraform using pagerduty_schedule and pagerduty_escalation_policy resources. Build primary and secondary schedules with explicit time zones and weekly rotations. Page the primary first, escalate to the secondary after 5 unacknowledged minutes, then escalate to the incident commander after another 10 minutes, and repeat the policy once. Attach the policy to the service, add a runbook URL to every triggering alert, support temporary schedule overrides for leave, and output a test procedure that proves each level receives a real notification.