The phone buzzes at an hour that should be illegal, and a half-awake engineer squints at an alert about a service that has fallen over. If this is the same engineer who answered the last three, something has gone wrong with the team long before the service did. A healthy on-call rotation makes sure the worst hours of operations are shared, and that what is learned at those hours is shared too. Here are eight lessons, told as small human moments, from teams that fixed the habit of leaving the pager with one or two heroes.
1. The hero is a warning light, not a trophy
Every team has someone who always picks up. They are proud of it, and the rest of the team is grateful. Look closer and you will see the cracks: shortened lunches, cancelled weekends, a quiet look when the phone rings at dinner. A person who is permanently on call is a sign the system is brittle, because the knowledge needed to respond lives in one person. Treat the hero as an early warning, and start planning the rotation while they are still smiling.
2. Put everyone in the pool, including the architect
A rotation that skips the most senior engineers teaches the team who counts and who does not. It also wastes the best education available. Those who designed the system learn how it behaves under stress only by sitting with the pager. One team lead told the story of a principal engineer who, on his first night in the rotation, discovered that a retry setting he had chosen years earlier turned small failures into storms. That lesson was worth more than a month of design review.
3. Pair the newcomer with a guide
Nobody should face their first shift alone. Let the newcomer shadow for a week, then lead with a guide listening in, then take a shift with the guide reachable. The guide is not there to take over, and rescuing too early robs the learner of the thing they came for. This is how the knowledge of a veteran travels into several heads instead of staying in one.
4. Make the schedule visible and swappable
Rotations break when they are hidden. Publish the calendar where all can see it, and let people swap openly. A swap made in the open is a kindness; a swap made in secret is a favour that someone eventually collects. Make the rule simple: you may trade, you must tell the team, and the pool stays the same size.
5. Build the handover into the rhythm
At the end of each shift the outgoing engineer leaves a short note: what fired, what was done, what is still odd. The incoming engineer reads it before anything else. Teams that do this for a few months find they have created a living history of their system without ever setting out to. When you are working on modern infrastructure, such as the kind described in this overview of handling very large traffic loads with cloud infrastructure, such a history is a quiet asset, because the weird behaviours of a busy platform are exactly what a note captures.
6. Protect the recovery
Rest is not a perk to be traded for dedication. Tiredness is cumulative, and people are poor judges of their own fatigue. The engineer who insists they are fine after three broken nights is usually the one most likely to type the wrong command on the fourth.
A rotation that ignores rest is simply a faster way to burn people out. After a rough night the next morning should be gentle, and a heavy week should be followed by a light one. Some teams give a half-day off after any night that included a page. It is not generosity for its own sake; it is how you keep good people in the pool for years.
7. Rotate the unpopular follow-up work, not just the pager
Alerts are only part of the burden. There is also the tedious clean-up afterwards: writing the incident summary, chasing the fix, updating the notes. If that also lands on the same two people, the rotation has only spread the glamorous part. Assign follow-up to the person who held the shift, and give them time to do it properly. This habit matters particularly for teams in the middle of a larger change, such as the shift described in this guide to moving a business from paper processes to digital platforms, because a transition produces more small incidents than a steady state does.
8. Ask what each person learned, every cycle
Close each rotation with one question for the group: what did you learn that you would not have learned otherwise? Answers vary from the technical to the personal. One engineer realised she could stay calm under pressure. Another found a monitoring gap that no dashboard had ever shown. The question keeps the rotation from becoming a chore schedule and turns it into a training programme.
Getting the first rotation moving
You do not need a new tool to begin. Take a sheet of paper and write the names of everyone who is technically able to respond, even if they feel unready. Next to each name, write the single thing they would need to learn first: access to the logs, an introduction to the deployment steps, a look at the escalation list. Then choose a start date a month out, and spend those weeks on shadowing. By the time the first fresh name takes a shift, there is a note waiting for them, a guide behind them and a calendar everyone has seen.
Expect two objections. Some will say the veterans are faster, which is true and beside the point; speed that depends on two people is a loan against the future. Others will say incidents are too serious to practise on. The answer is that they will happen anyway, and it is far better for a second person to meet the system for the first time on a calm Tuesday with a guide than on the worst night of the year with nobody.
Bonus: where outside help fits
Some teams are too small to build a fair rotation alone, or are short of people with the right experience. In that case a partner can fill in gaps rather than replace the team. DevOps and cloud operations consulting and support companies, such as DevOpsSupport and its consulting, support and training services, work with startups and larger organisations to build shared on-call habits and to cover stretches where the in-house team needs relief. The aim is to spread the load and leave the team more capable, not more dependent.
A small story to finish
Numbers and tooling matter less than people here, which is why the closing example is about a person.
One team we heard about began with a single engineer, Min, who had answered nearly every page for two years. When she finally told her manager she was thinking of leaving, the manager did not offer a raise first. He drew a grid on the whiteboard with six names. Within a quarter Min was on call one week in six, and, to her surprise, she slept better than she had in years. The other five engineers, who had always assumed incident work was beyond them, discovered they were good at it. Their first handover notes were clumsy. Their tenth were excellent.
If you are still the only one who answers, do not wait for a crisis to spread the work. Start a grid. Add names. Write the notes. The team that shares its hardest nights stays awake less and learns more, and that is the best trade there is.

