Understanding your numbers
How uptime and availability are calculated
What the dashboard percentage measures, why a site is only "down" when everything in it is, and which number to put in front of a customer.
Device uptime
A device’s uptime is the share of the last 30 days it was reachable. It is measured from recorded status changes, not from sampling — so a two-minute outage counts as two minutes even if it fell between polls on either side.
A device shows "tracking started" rather than a percentage until it has enough history. That is honest rather than flattering: a device connected an hour ago has not been up for 30 days, and printing 100% would imply it had.
Site uptime is not the average of its devices
This is the part worth understanding, because it is the one that surprises people.
| Site state | Meaning |
|---|---|
| Up | Every monitored device in the site is reachable. |
| Degraded | Some devices are down, but not all of them. |
| Down | Every monitored device in the site is down at the same moment. |
So a site with twenty access points and one dead one is degraded, not 95% down. A site is only counted as down when the whole thing is unreachable — which is what a site outage actually is, and what a customer means when they say "we were down".
That also means a site can read 100% up while a device inside it was offline for a day. The site figure answers "was the site reachable"; the device figure answers "was this box reachable". Use the SLA report when you need per-device numbers a customer can audit.
Dashboard availability versus the SLA report
| Dashboard | SLA report | |
|---|---|---|
| Window | Fixed at the last 30 days | Any period you choose |
| Maintenance windows | Not excluded | Excluded |
| Best used for | A live sense of how things are | Anything a customer will read |
If the two numbers disagree for the same period, planned maintenance is almost always the reason, and the report is the one to send.
How long history is kept
Raw status events are kept for 30 days. Before they are removed they are rolled up into daily figures, and it is those rollups that SLA reports read — so a report covering last quarter still works even though the underlying events are gone.
Retention is a floor, not an exact cut-off. A day is never purged before it has been rolled up, and a device’s most recent event is never deleted, so you may see raw events slightly older than 30 days. Nothing is lost early; occasionally a little is kept late.
Related
Did this miss something? Tell us — support is a small team and the fastest fix is usually a better article.