DevOps · Cluster
On-Call
Escalation, runbooks, and fatigue.
- 01 Escalation Policies as Timeout-Driven State Machines How an alert moves from first notification to a human who acknowledges it, what each timer actually measures, and where the routing logic breaks down.
- 02 Alert Fingerprints, Grouping, and Inhibition How an alerting system decides that two firing alerts are the same problem, how grouping batches notifications, and why unstable labels cause page storms.
- 03 Rotation Coverage Math and the Handoff Contract How 168 hours of weekly coverage divide across a team, what shift length costs, and what has to cross the boundary at every handoff.
- 04 What a Page Must Carry to Be Actionable The fields a notification needs so a responder can decide whether to act, where to look first, and who else is affected, within the length limits of a phone screen.
- 05 Anatomy of a Runbook Step Why a runbook step needs a precondition, a bounded action, and a verification that proves the system changed, and how steps fail when any of the three is missing.
- 06 Severity Classification as a Routing Input How a severity level selects the channel, the responders, and the update cadence, and why definitions anchored in observable impact outperform definitions anchored in alarm.
- 07 Measuring Alert Fatigue Instead of Describing It Which quantities make page volume and actionability comparable across rotations, how to collect them without rewarding the wrong behaviour, and how the numbers drive deletion.
- 08 Incident Roles and the Status Update Cadence Why the person with the most context should not be the one coordinating, how the commander, operations, and communications roles divide the work, and what a status update owes its readers.
- 09 Follow-the-Sun Coverage and the Cost of Every Handoff The arithmetic of covering a full day with regional shifts, the overlap a real transfer needs, and why an incident that crosses three shifts loses more than time.
- 10 From Postmortem to Alert Change: Closing the Detection Loop How a reconstructed timeline becomes a specific detection gap, how action items are classified so they change alerts and runbooks, and how to prove the change took effect.