In “Site Control,” availability checks are scheduled every five minutes. Each check shows the site’s response to a single request, not its status during every second between requests. A brief outage can start and end before the next check.
That’s why a visitor’s report of a problem and two successful monitoring results can both be accurate. To investigate, save the address and time of the visitor’s attempt, then compare them with the checks that actually ran.
Look at a two-minute outage between checks
Suppose the service requests the page at 10:00 and 10:05. A brief outage occurs between those requests:
| Time | What was happening with the site | What the check recorded |
|---|---|---|
| 10:00 | The page is available | Successful response |
| 10:02 | The outage begins | No request this minute |
| 10:03 | The visitor gets an error | Visitor report received |
| 10:04 | The site is responding again | No request this minute |
| 10:05 | The page is available | Successful response |
This is an illustrative timeline. It explains the limitation of periodic requests; it does not promise the exact time every real check will run. Two successful responses do not mean the visitor was mistaken.
In this situation, ask the visitor for the exact address and approximate time, and to save the error message. Send the time interval to the developer or hosting provider so they can check the server records—the logs. The monitoring guide explains the purpose of availability checks in more detail.
Don’t call the detection time the exact start of the outage
If the site responded at 10:00 and the check showed a problem at 10:05, you know the site’s status at those two moments. The data does not show exactly when the outage began between them.
The first successful response after a problem also does not establish the exact minute service was restored: the site may have started working earlier. In a report, it’s better to write “the check at 10:05 showed an error; a normal response was received at 10:10” than “the site was down for exactly five minutes.”
If a specialist determines the start and end times from the logs, include the source and time zone. For example: “Monitoring detected an error at 10:05 MSK; according to the server log, the outage began at 10:03.” Don’t substitute one timestamp for the other: they answer different questions.
Check the actual dates, especially when data is missing
Five minutes is the planned frequency, not a guarantee that every request will run on time. An individual check may be delayed or may not produce a result. The terms of the Site Control subscription should not be presented as a promise of continuous monitoring.
If the latest result is old, don’t assume the site was available the whole time just because its previous status was green. First open the address manually, check the next completed check, and find out why there is no fresh data.
In your records, distinguish between three cases: “normal response received,” “error received,” and “no new result.” The last case confirms neither that the site is working properly nor that it is currently unavailable. This keeps the report from concealing a period of uncertainty.
Collect a second source for recurring brief problems
Record complaints with the address, time, device, and exact action. If the problem only occurs when submitting a form, checking the homepage alone is not enough. The article about a broken checkout outlines how to check that user journey.
A short message to the hosting provider could say: “Visitors reported an error at /catalog at 10:03 and 12:18 MSK. Nearby monitoring checks returned successful responses. Please check this address and these time intervals against the logs.” If the time is approximate, say so.
Compare the information while keeping its differences clear. Monitoring checked the public address, the visitor clicked a button, and the server log may record a different step. Matching times alone do not explain the cause. The investigation should result in either a specific confirmed incident or an accurate record of what information is still missing.