A monitor checks at a fixed interval whether your service responds. That interval is the most important setting: it decides how long an outage can go unnoticed.
The arithmetic
If a service goes down right after a check, only the next check notices. In the worst case a whole interval passes, on average half of one. With a five-minute interval an outage can stay unnoticed for five minutes; with 30 seconds, for at most 30 seconds.
Short outages that start and end between two checks are never seen at all. If you need to catch those, you need a shorter interval.
Shorter is not always better
- More requests: every check is a real call to your service. For a heavy page, a light endpoint such as a dedicated status page is worth it.
- More alerts: by default Uptime reports a monitor as down after the first failed check; in the notification settings you can require several failed checks in a row. A service that occasionally drops a single request will trigger alerts more often with a short interval.
- Timeouts: a check waits a few seconds for an answer. An interval that is barely longer than that wait gains nothing.
A rule of thumb
- Shop, sign-in, checkout, public API: as short as possible, in Uptime one minute, on the Business plan also 30 seconds.
- Company website, blog, internal tools: one to five minutes.
- Services whose failure only shows after hours, such as nightly jobs: five minutes or more.
More important than the perfect interval is that someone who can act actually gets the alert. So check who the notifications go to as well.
