November 1, 2026
What is uptime monitoring? How it works and what to check
Your site can go down while your server dashboard looks normal. An expired certificate, a bad DNS change or a failed deploy can all stop your customers. Without a check from outside, the first alert is often a customer email.
Uptime monitoring is a service that requests your website or API from outside your network on a fixed schedule, records the result, and alerts you when a check fails. It answers one question: can people reach this service right now?
How an uptime check works
- The monitor requests your address on a schedule, for example every 30 seconds.
- It compares the answer with your rule: a status code, a response time or a word on the page.
- If the check fails, a good service tests again, often from a second location, to reject false alarms.
- It alerts you by email, chat or phone, and records the start of the incident.
- When checks pass again, it closes the incident and records how long the outage lasted.
You can try one check now. Our website down checker tests any address from 22 locations, with no account.
What you can monitor
Start with HTTP. Add other check types when a failure type has hurt you, or could.
| Check | What it finds |
|---|---|
| HTTP | The page returns an error or does not answer |
| Keyword | The page answers, but the expected text is missing |
| SSL | The certificate is about to expire |
| DNS | A DNS record changed or does not resolve |
| TCP and ping | A port or host does not answer |
| Heartbeat | A cron job did not run on time |
| Browser | A login or checkout flow breaks |
How often should you check?
The interval sets the longest time before you know about a problem. It also decides if you see a short outage at all. This table shows the result for an outage of 45 seconds.
| Interval | Longest wait for a first failed check | Outage of 45 seconds |
|---|---|---|
| 30 seconds | 30 seconds | Finds it |
| 1 minute | 1 minute | Can miss it |
| 5 minutes | 5 minutes | Often misses it |
| 30 minutes | 30 minutes | Usually misses it |
The interval also limits what you can measure. A 99.99% target allows about 4.38 minutes of downtime a month, so a check every 5 minutes cannot measure it well. See what a good uptime SLA is for each target.
Why one location is not always enough
A check from one location can fail because of a network problem near that location, while your site is fine. A check can also pass there while users in another region cannot reach you. More locations reduce false alarms and show regional outages. Read about our 22 check locations.
Uptime monitoring and related tools
| Tool | The question it answers |
|---|---|
| Uptime monitoring | Is it up from outside, right now? |
| Browser (synthetic) monitoring | Can a user still finish a login or checkout? |
| APM | Which code or query inside the application is slow or failing? |
| Status page | What do we tell customers during an outage? |
What uptime monitoring does not do
- It does not find the cause inside your code. You need logs or APM for that.
- A basic HTTP check does not prove that a login or checkout works. You need a browser check for that.
- It does not fix the outage. It only makes sure that you know about it early.
How to set it up
- List the addresses that matter: home page, login, API health check, checkout.
- Create one HTTP monitor for each address, at the shortest interval that you can get.
- Add a keyword rule to pages that can return 200 with an error message.
- Send alerts to a place that someone reads, such as Slack, Telegram or Discord, and also to email.
- Stop one test service, or point a monitor at a bad address, to confirm that the alert arrives.
Frequently asked questions
What is uptime monitoring in simple terms?
A service requests your website or API on a fixed schedule from outside your network. If the request fails, the service alerts you. It also records each result, so you can see your uptime over time.
How is uptime calculated?
Uptime is the time a service was up, divided by the total time in the period, times 100. A site that was down for 43.8 minutes in a 730-hour month had 99.9% uptime.
How often should a website be checked?
Use the shortest interval that you can get for pages that earn money or that customers log in to. A check every 30 seconds finds an outage within 30 seconds. A check every 5 minutes can miss a short outage completely.
Do I need to install anything?
No. HTTP, keyword, DNS, SSL, TCP and ping checks run from outside and need only an address. A heartbeat monitor is the exception: your cron job must send a request to a URL each time it runs.
What is the difference between uptime monitoring and APM?
Uptime monitoring tests your service from outside and tells you that it failed. APM runs inside your application and helps you find out why. Many teams use both.
Start with 30-second checks
The free plan has 5 HTTP or keyword monitors at 30-second checks from one location, with no credit card. Paid plans start at $3/mo with more locations and a 30-day money-back guarantee.
See plans