On-Call for a Small Team With No DevOps: A Simple Setup
Three developers and a live product, but nobody owns incidents? Set up a weekly on-call rotation, pick the few alerts that may wake someone, write one-page runbooks and keep the load fair.
A team of 2 to 5 developers can run on-call with a weekly rotation, one person on call at a time and a named backup, a short list of alerts that are allowed to page a human, and a one-page runbook for each. Start with a shared chat channel and your phone's notifications before you pay for a paging tool. Review every page after the week ends and remove or fix any alert that was not worth waking someone.
Most advice on this topic is written for companies with a dedicated platform team. If you are a founder or one of a few developers, you need something lighter that still means a customer is not left waiting while everyone assumes someone else saw the problem. This guide shows the smallest setup that works.
Step 1: Decide What Counts as an Emergency
On-call fails first because too many things page. Before you build a schedule, write down the few events that deserve a phone call at 3 a.m. For most small products that list is short:
- The site or API is down for real users.
- Sign-up, login or checkout is failing for many users.
- Payments are being charged but not fulfilled.
- The database is full, unreachable or out of connections.
- Data is being lost or leaked.
Everything else goes to a channel that someone reads in working hours: slow pages, a single user's error, a cron job that can wait until morning. The Checkly on-call guide and similar write-ups make the same point: alerts that page should need a human right now. If you already get a flood of alerts, fix that first with Too Many Production Alerts? How to Fix Alert Noise Fast.
Step 2: Pick a Rotation Length That Fits Your Size
Common patterns from small team guides and Hyperping's article on small teams:
| Team size | Simple setup |
|---|---|
| 2 people | Alternate weeks. The off-week person is the backup of last resort. |
| 3 people | One week each. Primary and backup rotate so everyone does both. |
| 4 to 5 people | Weekly primary, plus a backup. Each person is on call about one week in four or five. |
| Solo founder | Be the primary, and give a trusted friend or contractor a backup role for outages longer than an hour. |
Weekly shifts keep context: the person on call knows what changed during the week. Some tiny teams rotate daily and let people swap when they need a day off. Both work. What matters is that the schedule is written down and everyone can see who is on call right now.
Put the schedule where people already look. A pinned message in your chat tool with the name, the backup and the handover day is enough to start.
Step 3: Route Alerts to a Phone, Not Just a Channel
A chat channel nobody has notifications on at night is not on-call. Choose the lightest option that actually rings:
- Chat app with push notifications for the on-call person. Free. Works if the on-call person turns on notifications for the alert channel and mentions them directly.
- Email-to-SMS or a small paging tool. Many monitoring tools can send SMS or phone calls on their own.
- A dedicated paging service. Worth it when you have more than a handful of people or need escalation rules. See PagerDuty alternatives for lighter choices.
Whatever you pick, test it. Trigger a fake alert on a Monday morning and confirm that the primary's phone makes a noise. Then confirm that the backup gets the alert if the primary does not acknowledge it within 10 to 15 minutes.
Step 4: Write One-Page Runbooks
A runbook is a checklist for a known failure. It is what lets a tired person without deep infrastructure knowledge act in minutes. Write one for each alert that can page, each on a single page:
# Checkout is failing
**How you know:** alert "checkout error rate above X" or support message.
**First check (2 min):** open the status of the payment provider; check the last deploy time.
**If a deploy happened in the last hour:** roll back to the previous release (link to the steps).
**If the payment provider is degraded:** post to the status page, pause promotions, wait.
**Who to call next:** backup on-call, then the founder.
**After:** add notes for the postmortem.
Keep them practical: exact commands, dashboard links and the rollback step. A runbook that says "investigate the issue" is not a runbook. One common rule is that an alert without a runbook does not get to page anyone.
Step 5: Give the First Ten Minutes a Script
Most incidents are handled badly in the first few minutes, not the last. Agree on a simple order:
- Acknowledge the page so the backup does not also start.
- Check what changed. Did a deploy go out? Did a dependency change? Rolling back a recent release fixes more incidents than clever debugging.
- Tell users. Update your status page with one honest line, even if it is "We are looking into it."
- Fix or mitigate. Roll back, restart, scale, disable the feature flag.
- Escalate to the backup if you are stuck after about 15 minutes.
Do the deep investigation after users are back. A short written note of what you saw and did, taken while it happens, makes the later review much easier.
Step 6: Keep It Fair
On-call at a small company becomes a problem when it is invisible work. Options that teams use to keep it sustainable:
- Time off in return, such as a late start after a night page.
- A stipend or extra pay for the week.
- Swaps with no questions asked.
- No on-call during holidays, unless people choose it.
Also protect the person on call from other work. Their week should have fewer planned tasks, so a page does not push everything else behind.
Step 7: Review Every Page
After each shift, spend 15 minutes on a short review. For every page that fired, ask:
- Was it a real problem for users?
- Could it have waited until morning?
- Did the runbook help?
- Can we prevent it from happening again?
If one week has far more pages than another, the schedule is not the problem. The system is. Delete or downgrade alerts that did not need a human, and turn repeat incidents into fixes. For serious outages, use a short incident postmortem so the lesson is written down once.
How Relia Helps
Relia watches users, APIs, the database and servers of a live product and joins related signals, such as a deploy, a slow query and users hitting an error, into one incident explained in plain words. That gives the person on call one thing to read instead of five separate alerts, and it shows who was affected. When there is a stack trace it gives a fix plan with the file and line, and after a deploy it checks whether the fix worked. It also has uptime monitors, checks of key pages and a public status page. Relia has a Free plan, and Growth is $19/month. You can look at the live demo to see how an incident reads.
FAQ
How many people do you need for an on-call rotation?
You can start with two. Below that, it is one person carrying all the risk, so add a backup as early as you can, even a part-time one.
Do we need PagerDuty?
Not at the start. A chat channel with push notifications for the on-call person, or the SMS and phone alerts built into your monitoring tool, can be enough. Add a paging service when escalation and scheduling become hard to manage by hand.
How long should an on-call shift be?
Weekly is the most common. Daily shifts spread the load but lose context. Pick one, write it down and adjust after a month.
Should the on-call person be paid?
Many teams give compensation in the form of time off, a stipend or extra pay. It keeps the rotation fair and it reduces churn.
What if I am the only developer?
Reduce what can page to the true emergencies, automate restarts where safe, publish a status page, and find one backup person for long outages.