On-call for a team of five
Five is a floor for a sustainable rotation: one primary each week and a secondary as backup, so each person carries the pager about one week in five. It works only while page volume is low, handoffs are clear and leave is planned. When pages grow, fix the noise first, then hire or buy cover.
What five people can carry
Five engineers is a common size for a first real on-call rotation, and it is a floor, not a comfortable size. It is enough to share the load when pages are few, and small enough that the arithmetic is unforgiving. With one primary each week, each person is primary one week in five. Add a secondary who backs up the primary, and each person is also secondary one week in five. That is two weeks in every five with some duty, though the secondary’s load is lighter.
The arithmetic has a weak point. Take one person out for leave or illness and the rotation is four. Lose someone for good and the rest absorb it. A team of five is sustainable when the alerts are few, and fragile when they are not. The Being On-Call chapter of the Google SRE book gives its own guidance on how many people a rotation needs and how much work a shift can absorb. It is written for a busy service with a dedicated SRE team, so a small team with low page volume should read it as a caution about load, not as a rule to copy. The business reasons are simple: a rotation that burns people costs retention, and lost engineers cost more than the pager ever saved.
Rotation shapes
Teams of five usually pick one of a few shapes.
| Shape | How it works | Load on each person | Fits when |
|---|---|---|---|
| Weekly primary only | One person carries the pager for a week; a lead escalates | One week in five | Very few alerts; a lead who is reachable |
| Weekly primary and secondary | A second person backs up the first | One primary and one secondary week | The usual choice for a team of five |
| Weekday and weekend split | Different people for weekdays and weekends | Shorter blocks, more handoffs | Heavier weekends, or people who prefer it |
| Follow-the-sun | Each region covers its own daytime | Waking hours only | People in far-apart time zones, never one office |
The second shape is the default for good reason. The third helps when weekends are busy, at the cost of more handoffs. The fourth is often wanted and rarely possible: five people in one time zone cannot follow the sun, because somebody has to be awake at every hour.
Primary and secondary
The primary takes the first page. The secondary is the next step in the escalation policy: paged if the primary does not acknowledge within the timeout, and called in when a second problem overlaps the first. Keep the roles clear. The primary owns the incident, and the secondary supports and covers.
The secondary is also a training seat. A newer engineer can be secondary before becoming primary, so they see real pages with a safety net behind them. That grows the pool of people who can carry the pager, which is the most useful protection for a small team.
Handoffs
A rotation of five hands over every week, so the handoff is the moment risk enters. Do it in working hours, not at midnight. A good handoff is short and written:
- Open incidents and follow-ups. What is still unresolved.
- Flaky alerts. The ones likely to fire again, and what to do about them.
- Planned changes. Deploys, migrations and maintenance in the coming week.
- Known risks. Anything the outgoing person is worried about.
A fifteen-minute conversation beats a long document. Keep the note where the next person will find it, and keep it in one place.
An example week
This example is invented. A team of five runs a weekly primary and secondary rotation. On Monday morning the outgoing primary hands over in a fifteen-minute call: one alert is flaky and likely to fire again, a migration is planned for Thursday, and one incident still needs a follow-up. The incoming primary reads the note and asks two questions.
On Wednesday night the primary is paged, acknowledges, and finds a problem that needs a second pair of eyes. The secondary is brought in by the primary’s request, not by the policy, and the two fix it together. On Thursday the primary takes an afternoon off to recover, which the team agreed in advance, and the secondary covers the migration. On Friday, the review of the week’s pages shows that the flaky alert fired four nights running and cost the primary most of their sleep.
That last finding is the useful one. It goes on the list for the next noise review, because the cheapest way to make a rotation of five sustainable is to make it quieter.
Leave, illness and departures
Leave breaks small rotations. Three habits make it manageable.
- Publish the rota weeks ahead. People plan holidays around it, so a late change costs goodwill.
- Make swaps easy. A swap should take one message and one change to the schedule, and both people should agree.
- Name a reserve. A lead or manager who steps in when someone is unwell, so a gap never has to be filled by the person who has just had a bad night.
Never schedule someone on call during their leave, and write down the plan for a team of four. It is the plan that matters when someone leaves.
Fairness, and what it costs the business
A small team notices unfairness fast. Count the load, not just the weeks: how many nights each person was paged, how many weekends and holidays they covered. Share the awkward slots evenly, ramp new joiners in as secondary first, and do not let one senior engineer become the answer to every hard page. How to pay for the burden is covered in on-call compensation.
The cost to the business is easy to miss because it arrives late. An engineer who is woken often is less effective the next day. Someone who feels the load is unfair starts to look for another job. Customers feel both, in slower fixes and in turnover on the team that knows their systems. A fair, quiet rotation is a retention tool as much as an operational one.
When to hire, or buy cover
Stretching has a limit. Signs it has been reached: the same few people cover most nights even after noise has been cut; leave cannot be covered without overloading others; or customers expect a faster response than five people can give. There are four responses, and they can be combined.
| Response | What it does | The trade-off |
|---|---|---|
| Cut the noise | Fewer pages for the same team | Takes effort first, but pays back fastest |
| Hire a sixth engineer | One week in six, not one in five | Slow, and a real cost |
| Add a time zone | Someone awake for the hours that hurt most | Coordination cost, and hiring in another place |
| Buy first-line cover | An outside team triages and escalates | It costs money, and needs clear runbooks to work well |
Start with the first row. A 30-day alert fatigue plan often gives a small team more relief than a new hire, at lower cost and sooner. Hire or buy cover when the pages are real, the noise is already low, and the team is still stretched. If buying cover or changing tools is part of the answer, plan the switch with moving your on-call without a missed page.
Frequently asked questions
Is five people enough for 24/7 on-call?
Yes, if the alert volume is low and everyone works in a compatible time zone. Five people give each person one primary week in five. It is tight, because leave, illness and departures each remove someone from the rotation, so the plan must say what happens when the team is four.
Do we need a secondary on-call?
It is strongly worth having one. A secondary catches the page when the primary does not answer, and shares the load when two problems overlap. The secondary role can be lighter, and it is a good place to bring in newer engineers who are learning the systems.
Can a team of five do follow-the-sun?
Only if the five are spread across time zones far enough apart to cover the hours between them. Five people in one location cannot. Follow-the-sun needs people awake in each region, which usually means at least two or three people in each of two or more zones.
How do we cover leave and illness?
Publish the rota weeks ahead, make swaps easy, and name a reserve, such as a lead or a manager, who steps in when someone is unwell. Do not schedule a person on call during their leave. The rota for four people is the plan that matters, so write it down.
When should we stop stretching and hire?
When the same people cover most nights even after noise has been cut, when leave cannot be covered without overloading others, or when customers need faster response than the team can give. Adding a sixth person changes the rotation from one week in five to one in six.