The On-Call Handbook: Rotations, Escalations, and Burnout
Healthy on-call programs need fair rotations, clear escalation, actionable alerts, runbooks, incident reviews, and active burnout prevention.
On-call should be designed, not endured
An on-call rotation protects customers, but it can damage teams when alerts are noisy, ownership is unclear, or engineers are expected to be heroic every night.
Good on-call design starts with respect for attention.
Rotations and escalation
Rotations should be predictable, documented, and fair. Primary responders need a secondary escalation path and a clear rule for when to page help. No one should be stuck deciding alone whether a database outage, security concern, or enterprise-impacting incident needs leadership involvement.
Escalation policies should match severity. Not every warning needs a page. Customer-impacting downtime, failed critical workflows, and data integrity risk deserve urgency.
Burnout prevention is reliability work
Reduce alert fatigue by removing duplicate alerts, tuning thresholds, adding context, and routing notifications to the right owner. Runbooks should be current and short enough to use at 2 a.m.
After incidents, review not only the technical failure but also the human load. A healthier on-call program improves uptime because rested engineers make better decisions.