Infrastructure & Operations › Working in Production

On-Call Handover

Passing open issues and context to the next person on call.

Also known as: on-call handover, shift handover, handoff

An on-call handover is the pass from one on-call engineer to the next at the end of a shift. Without it, the incoming person starts blind: they don’t know what’s shaky, what’s been “fixed” but isn’t confirmed, or what’s about to happen. A short, written handover turns a cold start into a warm one.

What to include:

• Active or recent incidents — anything open, anything only just mitigated, and what’s still being watched.
• Fragile or suspect systems — the service that’s been flapping, the alert that fired twice, the box that looks wrong.
• Changes and events coming up — a maintenance window, a big launch, expected traffic, a scheduled job.
• What’s already been tried — so the next person doesn’t repeat work or undo a mitigation.
• Anything “small” that isn’t worth a page but is worth knowing.

Handover — Tue 18:00, outgoing: Sam → incoming: Priya
- API latency elevated since 15:00; mitigated by scaling, watching.
- DB replica 2 restarted; monitor for lag.
- Maintenance window Thu 02:00 for storage upgrade.
- Checkout flag "new-cart" ramping to 50% tomorrow.

The classic mistakes:

• No handover at all. The incoming engineer rediscovering everything from scratch wastes the first hour of their shift.
• Verbal only. A spoken “you’re all good” leaves nothing to refer back to. Write it down — a shared doc or the on-call tool — so both people and history have it.
• Hiding the messy parts. Mentioning only the clean items and leaving “small” worries unsaid sets up the next person to be surprised.
• Not flagging upcoming changes. A launch or maintenance window the next person doesn’t know about is an incident waiting to happen.

Good handovers reduce toil and shorten incident response, because context travels with the shift. Keep it brief and factual, and link out to the runbook and open tickets rather than pasting their contents. Frontend, backend and data shifts all hand over the same way: what’s on fire, what’s fragile, what’s coming.