Checked once a minute from outside our primary region. A minute counts as up when model requests and the dashboard both answer within 10 seconds.
A separate probe on Cloudflare runs the same checks once a minute, so a failure that takes down AWS as a whole still shows here.
Uptime no data over the minutes measured since the probe started.
Per UTC calendar month, the figure the SLA uses: a minute is down when either probe saw it down. Down and measured are in minutes. A month is partial while it is under way or when the probe started inside it.
No month measured yet.
No incidents in the last 90 days.