How checks work

A check

A check is one look Mon takes at a monitor: a request to a website, a connection to a port, a reading of a certificate or a domain's registration. Each monitor is checked on its own schedule, and every check is recorded with what it found and how long it took.

How often

You set each monitor's interval when you add it - from every minute up to once a day. New monitors start at every 5 minutes.

Monitor type	Most often, on paid plans
HTTP, Keyword, Ping, Port, DNS Record	Every minute
Trace Route	Every 5 minutes
Synthetic	Every 15 minutes
Defacement, Redirect	Every hour
SSL, WHOIS, Expiration, DNS Zone	Once a day, always

On the Free plan, no monitor is checked more often than every 5 minutes.

SSL, WHOIS, Expiration and DNS Zone monitors are always checked once a day - a certificate or a domain registration changes rarely, so there's nothing to gain from more.

Where checks run from

Mon checks from 7 locations on 4 continents: Virginia, California, Dallas, New York, São Paulo, Frankfurt and Singapore.

By default a monitor uses Automatic Location, which lets Mon use every location and spread the work itself - the recommended choice. Turn it off to pick the locations yourself; at least one has to stay selected. The IP addresses of every location are listed in your account, under Locations, ready to allow through a firewall.

WHOIS, Defacement, Redirect, Expiration and Synthetic monitors always choose their locations automatically.

How Mon decides up or down

A single location having trouble shouldn't wake you up. So a check doesn't rely on one place:

1. Mon sends the check to one of the monitor's locations, favouring the ones that have been most reliable lately.
2. If it succeeds, the monitor is up, and the check ends there.
3. If it fails in a way another location might not see - a timeout, a refused or reset connection, a DNS lookup that failed, an error status code - Mon tries the next location, and so on, up to 8 of them.
4. Only when every location tried has failed is the monitor down.

Some failures look the same from anywhere, so Mon doesn't try elsewhere for them: a certificate problem, a keyword that's missing (or present when it shouldn't be), a redirect loop, a page that ends up at the wrong address, a domain that's about to expire.

During a long outage, Mon tries fewer locations per check, down to one, so it isn't sending a burst of requests to a site that's already struggling.

Down, up and events

When a monitor's state changes, Mon records an event: Up, Down, Paused, and Started monitoring when a monitor is first added. Each event stays open until the next one, so you can see how long every outage lasted and why it happened.

A monitor goes down on the first check that fails everywhere it was tried. Whether that alerts anyone straight away is up to each contact - see Alerts and contacts for alerting only after several failures in a row.

A monitor's figures

A monitor's page opens with its figures, as of its last check:

• Response Time - the average over the last 24 hours.
• Uptime - the share of the last 30 days it was up.
• Apdex - how many of the last 24 hours' checks answered quickly enough, as a score from 0 to 1.

Each one shows how it moved against the period before it - the 24 hours before, or the 30 days before - in green when it got better and red when it got worse. A faster response time counts as better. A monitor with no checks in the period before has no comparison to show yet.

Apdex sorts every check into one of three:

• Satisfactory - answered within half a second.
• Tolerable - answered within 2 seconds.
• Frustrating - slower than that, or failed.

The score counts each satisfactory check as 1 and each tolerable one as a half, divided by the number of checks. It's rated Excellent from 0.93, Good from 0.84, Fair from 0.69, Poor from 0.49 and Unacceptable below that. Under Monitors, a filter shows only the monitors whose Apdex says they're answering slowly. Redirect monitors have no Apdex: a redirect is followed in a browser, which takes seconds, and that would be judged as slow.

Timeouts

The timeout is how long a check waits for an answer before counting it as a failure. You can choose anything from 10 milliseconds to 30 seconds; new monitors start at 10 seconds. SSL, WHOIS, Expiration and DNS Zone monitors always allow 30 seconds.

Pausing and maintenance windows

Pausing a monitor stops its checks until you resume it. When it resumes, its first check sets it up or down without sending any alerts, so resuming a monitor never raises a false alarm.

Maintenance windows are scheduled times when a monitor isn't checked - once, every day, every week or every month - so planned work doesn't page anyone. Add them under Maintenance Windows, then attach them to the monitors they cover. WHOIS and Expiration monitors don't use them.

What Mon keeps

• Every check for 30 days, with its result and timings.
• Hourly and daily summaries of response times and uptime for 90 days.

Resetting a monitor's statistics clears its checks, uptime and events but keeps its settings and alerts. Deleting a monitor removes its history with it.