A good on-call handover covers six things: open incidents, half-done work, silenced and noisy alerts, risky events in the coming rotation, who owns each open item, and proof that paging and access work for the person taking over. Write it down, talk it through for a few minutes, and have the incoming engineer test the pager before they accept the shift. The checklist below gives you the exact items, a note format you can copy, and the traps that make the first night of a rotation the worst one.
TL;DR
- Hand over in three parts: a written note from the outgoing engineer, a short live conversation, and a test page to the incoming engineer.
- The note must list open incidents, what has been tried, next update times, unfinished work with an owner, and every silenced alert with its expiry.
- Name the noisy alerts and explain why they are not real. This is the cheapest way to protect the next person’s sleep.
- List risky events in the coming rotation: deploys, migrations, vendor maintenance, certificate renewals, peak traffic days.
- Keep personal data out of handover notes. Link to tickets instead of pasting customer details.
Table of Contents
- What a handover has to carry
- A handover note you can copy
- Talk it through: the live conversation
- Test before you accept
- Handing over the noise: fewer interruptions, less burnout
- Small teams, solo engineers and time zones
- Keep personal data out of handover notes
- After the handover: follow-up and postmortems
- Make the handover easier with Uptime Beacon
- FAQ
- Sources
What a handover has to carry
Most handover failures are not dramatic. The outgoing engineer knows something, assumes it is obvious, and leaves. Then the incoming engineer rediscovers it at 3 a.m. One on-call rotation guide describes this as a tax: teams meet the same flaky alert in rotation after rotation because nobody wrote it down.
A handover checklist for rotation changes lists the essentials as open incidents, half-done work, silenced and known noisy alerts, risky events in the coming rotation, ownership of open items, and checks that paging and access work. We use those six as the spine and add detail to each.
1. Open incidents
For every incident that is still open, the note should give its status, who is leading it, when the next update is due, and what has already been tried. That last item matters most. Without it, the new engineer repeats the diagnosis you finished two hours ago. The same checklist recommends recording exactly these points: status, lead, next update time and what has been tried.
The GitLab handbook on handoffs adds useful detail: active issues, their impact, investigation steps, blockers and recommended next actions. Also record issues that were resolved during your shift and any follow-up they need, because a “resolved” incident often has a fix that is only a workaround.
2. Half-done work
This is the category people forget. A rollback you started and paused, a config change applied to one region only, a ticket you meant to file. Write each one with a plain statement of the current state (“applied in staging, not in production”) and the next step. If it cannot be finished, say who owns it now.
3. Silenced and noisy alerts
For each silenced alert, write what it is, why you silenced it, and when the silence expires. A mute with no expiry is a future outage nobody hears. For noisy alerts that you left active, GitLab’s guidance is to record which alerts fired, why they were not real, and what to do when they fire again. The incoming engineer then reads “disk warning on the batch host, clears itself after the nightly job, ignore unless it persists past 6 a.m.” instead of treating it as an emergency.
4. Risky events in the coming rotation
List what is scheduled that could page the new person: deployments, database migrations, supplier maintenance, certificate renewals and peak trading days. Include the time, the owner and what to watch. If a vendor is doing maintenance on Thursday night, the incoming engineer should already know that failing checks that night are expected, and whether a maintenance window has been set up.
5. Ownership
Every open item gets a name next to it. “The team” is not a name. If an item belongs to someone off-rotation, say how to reach them and when it is acceptable to wake them.
6. Paging and access
The incoming engineer should be able to confirm, before the shift starts, that their phone rings, that they can log in to the systems they need, that the runbook links open, and that escalation contacts are current. More on that below.

A handover note you can copy
Keep the note short enough that someone reads it on a phone. Here is a format that fits one screen.
| Section | What to write |
|---|---|
| Open incidents | Status, lead, next update time, what has been tried |
| Resolved this shift | What broke, the fix, any follow-up still owed |
| Half-done work | Current state, next step, owner |
| Silenced alerts | Alert name, reason, expiry |
| Noisy alerts still active | Why it is not real, what to do if it persists |
| Recent changes | What changed, when, why, what to watch |
| Coming up | Deploys, migrations, vendor maintenance, cert renewals, peak days |
| Contacts | Escalation path, backup engineer, vendor support |
| Links | Runbooks, dashboards, status page, incident channel |
The “recent changes” row comes straight from the GitLab guidance, which asks for changes, when they occurred, why they were made and what to watch afterward. A change made on Tuesday is the first suspect when something breaks on Wednesday, so make it easy to find.
One rule of thumb: if a line in the note is something you would tell the person at the door while handing over the laptop, it belongs in the note. If it is only background, link to it.
Talk it through: the live conversation
A written note does not answer questions. A short call does. The GitLab handbook suggests reviewing the handoff notes in a 15 to 20 minute synchronous conversation when possible, and sharing them asynchronously in Slack or a shared document when it is not. The checklist guidance describes the same three-part pattern: a written note, a short live conversation for questions, and a test page before acceptance.
Use the call to cover what writing handles badly:
- Which open item worries you most, and why.
- What you would do first if the quiet alert fires.
- Anything that feels off but has no evidence yet. A hunch about a slow database is worth ten seconds.
Keep the call to the point. If it runs past 20 minutes, the note was not clear enough, and that is the thing to fix.
Test before you accept
The handover is not complete until the incoming engineer has proved they can be reached and can act. Run this short test:
- Send a test page and confirm the phone rings, including when it is on silent or in a focus mode.
- Confirm login to the monitoring tool, the cloud console, the deploy system and the incident channel.
- Open every runbook link in the note. Dead links are a handover bug.
- Check the escalation contacts: names, numbers and who is secondary.
- Confirm the incoming engineer knows how to start a maintenance window if planned work needs one.
This takes five minutes and prevents the most embarrassing failure there is: the page that fires and reaches nobody.
Handing over the noise: fewer interruptions, less burnout
Burnout on call rarely comes from one big incident. It comes from broken sleep and from pages that did not need a human. We could not find a rigorous study to cite on exactly how handovers change burnout, so we will keep this to practice rather than numbers: a handover that names the noisy alerts gives the next engineer the context to ignore what is safe to ignore and respond fast to what is not.
Three habits help:
- Write down every false alarm you saw. Note the alert, the time, and why it was not real. Two entries of the same alert in consecutive notes means it is time to fix the alert instead of handing it on again.
- Give silences an expiry. Temporary mutes should come back by themselves, with a reminder in the note.
- Fix the source of the noise. Most false pages from external checks come from a single failed request, a network blip between one probe and your server. Requiring several failures in a row, confirmed from a second location, removes that class of page entirely. That is how Uptime Beacon opens incidents on paid plans, and it is the reason we wrote about confirming failures at two locations and about paging only for real failures.
The less noise in the system, the shorter the “known noisy alerts” section of your note gets. That is the goal: a handover that is mostly about real work.
Small teams, solo engineers and time zones
The checklist scales down. If you are two people swapping weeks, you can still do the written note and the five-minute test. If you are one engineer covering for yourself, the note is for future you: write down silences and in-flight changes before you sleep, because tomorrow’s you will not remember them.
For teams across time zones, the asynchronous path works if the note is disciplined. Post it in the same channel every time, in the same format, so the incoming engineer knows where to look. When there is any overlap at all, hold the short call. When there is none, ask the incoming engineer to reply to the note with questions and an acknowledgment, so the handover has a visible end.
For a new engineer’s first shifts, the guidance recommends a shadow or a named backup. Pair them with someone who has done the rotation, and say in the note who that person is and when it is fine to call them.
Risky events across regions
If your customers are spread out, “peak day” differs by region. A promotion in the US evening is a night shift for an engineer in Europe. Put the event in the coming-up list with the time zone stated, not just the hour.
Keep personal data out of handover notes
Handover notes live in chat history and shared documents, which means they are logs of a kind. Under the GDPR, personal data in logs should be limited to what is necessary, protected against unauthorized access, and deleted or anonymized under an established retention policy. The practical version for handovers:
- Refer to tickets and incident IDs. Do not paste customer names, emails or session details into the note.
- Post in a channel with restricted membership, not a company-wide one.
- Set a retention period for the handover channel or document and review it on a schedule.
If you operate in the EU or serve EU customers, where your monitoring results are stored matters too. Our piece on EU data residency questions goes beyond server location.
After the handover: follow-up and postmortems
The resolved-incident section of the note should end in a postmortem, not a shrug. Use a short, consistent format so people actually read it: our incident postmortem template is a copy-paste layout built for that. For the first hour of an incident itself, the incident response playbook maps the steps.

It also pays to look forward. Before a risky rotation, a team can run a pre-mortem technique: imagine the rotation went badly, and list why. The answers (an expiring certificate, an untested rollback, a runbook nobody has opened) go straight into the “coming up” section of your note.
Watch expiring certificates in particular. They are predictable, they page at the worst time, and they belong on the risky-events list well before the date. See why certificate expiry needs a second location.
Our perspective
We run production for a living, and our founders have spent more than 30 years combined doing it in banking, fintech, telecom and SaaS. The pattern we keep seeing is that handovers fail on what is not written, and that most of what is not written is noise. People do not document the flaky alert because they have stopped seeing it.
That is why we built Uptime Beacon to remove the noise at the source. An incident opens only after several failures in a row, confirmed from a second location on paid plans, so the person taking over is not inheriting a pile of network blips. The handover can then be about real work.
We also think the handover should be easy to interrogate. Every plan, Free included, lets an MCP-compatible assistant such as Claude, ChatGPT, Cursor or VS Code read your monitors, incidents and check history, and diagnose a failing monitor region by region. An incoming engineer can ask what failed overnight and where, instead of scrolling a channel. With “Allow changes” turned on for a team, the assistant can also pause monitors, acknowledge incidents and start maintenance windows. Nothing can be deleted over MCP.
Aaditya Parashar
Make the handover easier with Uptime Beacon
If you are building a rotation from scratch, the Free plan covers the basics at €0 per month with no card: 5 monitors checked every 5 minutes from EU Central, one alert channel (email or Slack) and one status page. Paid plans start at €14.99 per month for Starter, which adds maintenance windows, the tool you will want when the coming-up list includes planned work. Pro is €39.99 per month and Business €89.99 per month. See Uptime Beacon plans for the details.
![]()
For cron jobs and APIs, which are often the half-done work that falls through a handover, see how API and cron monitoring for SaaS teams works with heartbeat checks.
FAQ
How long should an on-call handover take?
Plan on a 15 to 20 minute conversation when you can hold one live, per the GitLab handbook’s guidance, plus a few minutes to test paging and access. The written note should be readable in a couple of minutes. If the call keeps running over, the note needs to be clearer.
Can a handover be done asynchronously?
Yes. The GitLab handbook says notes can be shared in Slack or a shared document when a live conversation is not possible. Use the same format each time, ask the incoming engineer to reply with questions and an acknowledgment, and still run the test page.
What should I do about alerts that fire but are not real?
Record each one in the note with the reason it is not real and what to do if it persists. If the same alert shows up in two notes in a row, fix or retire it. Requiring repeated failures confirmed from another location, as Uptime Beacon does on paid plans, removes many network-blip pages before they reach anyone.
Is a handover checklist the same as a shift handover template?
They overlap. A checklist tells you what to cover and test. A template gives you the layout to write it in. The table above is both: use its rows as the checklist and copy it as your template.
Sources
- Handoffs and Continuity, The GitLab Handbook
- On-Call Handover Checklist for Rotation Changes, CheckFlow
- On-Call Rotation Best Practices: SRE-Style Schedules
- Regulation (EU) 2016/679, EUR-Lex
- The Pre-Mortem Technique, FluidOps Blog



