How it works
What happens between your site failing and your phone buzzing
Four steps, and the third is the one most tools get wrong.
- Step 01
You add a URL
Give it a name and choose how often to check — anywhere from every minute to every hour. The first check runs within a minute, so you know straight away that it is working.
- Step 02
We check it, on schedule, forever
Every check records whether the site answered, what status code it returned and how long it took. Nothing is sampled or thrown away — the full history is there.
- Step 03
Two failures in a row means down
One failed check is usually a network hiccup, not an outage. We wait for a second consecutive failure before declaring anything, which is what keeps our alerts worth reading.
- Step 04
You get one email
Not one per failed check — one per outage, with the cause and the time in your own timezone. When the site recovers, you get exactly one more telling you how long it was down.
A real outage, minute by minute
A monitor on the default five-minute interval. Nothing here is idealised — this is the sequence the system actually runs.
09:00
Check passes
200 OK in 214 ms
09:05
Check passes
200 OK in 231 ms
09:10
First failure
No response within 10s. Recorded, but nothing sent — one miss is not an outage.
09:15
Second failure — outage declared
Email sent: what failed, and the time in your timezone. The downtime clock starts here.
09:20 – 09:35
Still down
Checks continue and are recorded. No further email — you already know.
09:40
Recovered
One more email: back up, down for 25 minutes.
Two emails for one outage. That is the whole of it.
What counts as down
Five distinct failures, and the email names which one it was. Knowing a site is down is half the information; knowing it is a DNS failure rather than a timeout is the half that tells you where to look.
Timeout
No response before your timeout — 10 seconds by default, up to 30.
Connection refused
The server is reachable but nothing is listening.
DNS failure
The hostname no longer resolves. Often an expired domain.
TLS problem
An expired, mismatched or untrusted certificate.
Wrong status code
Anything other than the code you expect. Configurable — a redirect page is healthy at 302.
Why we wait for two failures
A monitoring tool that cries wolf is worse than no tool at all. If you are woken at 3 AM for a blip that fixed itself in five seconds, you learn to ignore the emails — and then you miss the real outage.
So a single failure is recorded but not announced. It takes two in a row before we call your site down. The cost is that detection can be one check-interval slower; the benefit is that when an email arrives, it means something. If you want faster detection, set a shorter interval.
Choosing an interval
Detection takes two consecutive failures, so realistic worst-case detection is roughly twice your interval. Pick the row that matches what the site is worth to you.
| Check interval | Worst-case detection | Suits |
|---|---|---|
| Every 1 minute | ~2 minutes | Fastest available |
| Every 5 minutes | ~10 minutes | Default |
| Every 15 minutes | ~30 minutes | Low-traffic sites |
| Every hour | ~2 hours | Slowest available |
Two deliberate limits
Public addresses only
We refuse private and internal addresses, and we re-check that on every single request rather than once when you add the monitor. A hostname that resolves publicly today can point somewhere internal tomorrow.
One check location, email only
Checks run from a single region, so a purely regional outage may not be visible to us. Alerts are email — no SMS or Slack yet. We would rather say so here than have you discover it during an incident.
Common questions
- How quickly will I know?
- With a one-minute interval, typically within two minutes of the site going down. With the default five-minute interval, within about ten.
- Will you email me every five minutes while it is down?
- No. One email when it goes down, one when it comes back. That is it.
- What counts as down?
- No response before your timeout, a connection refused, a DNS failure, a TLS certificate problem, or a status code other than the one you expect. That last one is configurable — a page that always redirects is healthy at 302, not 200.
- Can I monitor something on my own network?
- No. We only check publicly reachable addresses, and we deliberately refuse private and internal ones. Our servers can reach places you may not intend, so that restriction protects you as much as us.
- How often can I check?
- Anywhere from every 60 seconds to every hour. We do not allow anything faster than a minute: at that rate a monitoring tool becomes a denial-of-service tool pointed at your own site.
- What if I am on holiday and a site flaps all day?
- You get one email per outage, not one per failed check, so a site that fails and recovers repeatedly cannot flood your inbox mid-outage. Every outage is still recorded and visible in the dashboard.