What it contains
The current state in words as well as colour, the affected parts, incident updates with times and a next-update promise, a history of recent incidents and a way to get in touch.
Good practice
Host it away from the service it reports on. Use a small fixed vocabulary (investigating, identified, monitoring, resolved), the four statuses Atlassian’s Statuspage documentation defines: investigating means symptoms are seen and the cause is unknown, identified means the root cause is found and a fix is under way, monitoring means the fix is believed to be in place while symptoms subside, and resolved means the root cause has been eliminated. Publish confirmed incidents only, so one failed check does not appear. Never colour a day with no data as green; unknown is not up. Keep it current, and close every incident with its duration. Google’s SRE book gives one person the communications role, whose duties include issuing periodic updates to the response team and stakeholders.
What not to publish
Customer data, internal blame, and a guess about the cause before it is identified. Say what is known, what is not known yet and when the next update comes. A promised fix time that passes in silence does more harm than “we do not know yet, next update at 14:30”.
For agencies
A status page per client reduces enquiries during incidents and gives evidence for the care report. It is worth offering where visitors or staff ask whether the site is down, not to every brochure site.
An example update sequence
14:07 Investigating: some visitors cannot load the shop. Next update by 14:30. 14:12 Identified: a DNS change sent traffic to the wrong address. 14:19 Resolved: the record is corrected, duration 12 minutes. Four lines with times, a fixed vocabulary and a promise of the next update answer most “is it down?” messages, and the closed entry is the evidence for the monthly report.
What to check
Who posts updates, who approves them out of hours and what the page shows when there is no measurement.
Sources
Read on . Standards and vendor pages change; confirm the current text before you rely on it.