Skip to content

How uptime is counted

The uptime figures on a monitor's page (Last 24 hours and Last 30 days), the Uptime (30 days) column of the Monitors list and the account's line above the list (such as 99.98% uptime, last 24 hours) are all worked out the same way, from the results of each check.

The share of results that found the website up

Every result counts as one of four things:

A result that is Counts as Examples
Up Up Up
Needs attention Up Slow, Access refused (403), Behind a sign-in page, the first 10 minutes of WordPress maintenance mode
Down, inside a confirmed outage Down Server error (500), Unreachable, Keyword missing once a second place agreed
Not counted Neither Could not check, failures a second place did not confirm, results during a maintenance window

The uptime is the up results divided by the up and the down results together. Results that are not counted are left out on both sides, so they neither raise nor lower it. Every result of the check results table says what it counts as.

Needs attention counts as up

A result that needs attention means the website answered, but not quite as it should: visitors may well get it. So it counts as up in every figure, and shows as an amber, striped bar so you still see it (my monitor says Needs attention).

Down only inside a confirmed outage

A failure counts as down only from the moment the monitor was confirmed Down, when a second place agreed (or one place found it failing 3 times in a row when no second place could check), until it is Up again. A failure that a second place did not see is not counted: its hour shows a grey bar, Not counted as down (down only when a second place agrees).

Not counted

  • Could not check: our checking server had a problem. It is never your website's fault, so it never counts.
  • Failures a second place did not confirm, shown as Checking again.
  • WordPress's maintenance page after its first 10 minutes, while the monitor is not confirmed down (WordPress maintenance mode).
  • Every result taken while the monitor was in a maintenance window, whatever it found.

Maintenance windows are left out

Results taken during a maintenance window count neither as up nor as down, so planned work never lowers your uptime. The time a monitor was in a window is recorded when it happens: editing or deleting the window later, ending it early, or moving the monitor to another group never changes the uptime that was counted before.

Never rounded up

The figure is cut to two decimals, never rounded up. A monitor shows 100% only when not one result in the period was down: one short outage in 30 days is enough to show 99.99% or less. No data means there is no counted result yet.

Which hours

Figure The hours
Last 24 hours The current hour and the 23 before it
Last 30 days, Uptime (30 days) The current hour and the 719 before it
The account's line above the list The same 24 hours as the bars, for every monitor except the paused ones

The Last 30 days figures are the same on every plan. A paused monitor has no results while it is paused, so those hours are left out. Each hour bar, when you point at it, says the share of that hour, such as Down about 15 minutes, 75% up.

Outages

Last 24 hours and Last 30 days also say how many outages there were. Hours in a row with a down result are one outage; one or two hours without a result in between, as a check every 30 or 60 minutes can leave, do not split it.

Which checks count

The figures come from the monitor's switched-on Uptime check. A port monitor counts the same way, from its switched-on Port check: its hour bars, its Uptime (30 days), its Last 24 hours and Last 30 days, and its share of the account's line above the list. A port that is closed, does not answer or says the wrong greeting counts as down once a second place agrees, as for a website.

Basic Auth off, or a Basic Auth password needed or refused, needs attention and never lowers any uptime figure: the website answers.