Write an on-call runbook for the pipeline only you know how to fix

You're the only person who knows how to fix pipeline X when it breaks, and you just went on vacation while it broke.

IntermediateRead 6 minDoing: a couple of hours, plus whatever time your teammate's test run takes7 steps

▶ Start

What you'll have at the end

A runbook a teammate follows successfully during a real incident, without paging you

Before you start

You need

A pipeline that breaks in one specific, recurring way you already know how to diagnose and fix, even though the fix has never been written down.

Not covered

This does not cover mapping the pipeline's stages or adding the observability checks that would catch the failure sooner; it only covers writing down the fix for one failure you already know how to handle.

Leans on

7 steps

Where to go next

An inherited pipeline, mapped stage by stage

if you don't yet know this pipeline well enough to name its one real failure and fix, map it first; this recipe assumes you already do.

Checked 26 Aug 2026

Part of the Data Engineering Core cookbook