An incident tells readers something is wrong. It shows on the page at once, changes the state of the components it affects, and can email the people who follow them.
Open it
Open Incidents and choose New incident.
- Type the Title: what readers see first, in a few words, like "Slow sign-in".
- Under What it affects, tick each component and choose the state the incident puts it in: Degraded, Partial outage or Major outage.
- Choose the Stage: Investigating while you look for the cause, Identified once you know it, Monitoring once a fix is out.
- Write the First update: what readers are living, in plain words. It takes Markdown. You can add more as you learn.
- Leave Email subscribers ticked to email the people who follow everything or one of the ticked components. Untick it for something small: that incident is then never emailed.
- Choose Open the incident.
Keep readers posted
On the incident's page, under Post an update, say What changed, move the Stage forward or keep it, and set the States after this update: a component that has recovered can go back to Operational while the incident stays open on the others. Choose Post the update.
A typo in an update can be fixed with Edit: the time stays, the update shows "edited", and nobody is emailed again. An update is never deleted.
Resolve it
Under Resolve, write a Closing note and choose Resolve the incident. Every component it held returns to Operational, and the subscribers who were emailed get the note.
If the problem comes back within 24 hours, choose Reopen on the incident's page: it returns at the stage and with the states it had. After 24 hours, open a new incident.
Explain what happened
Once it is resolved, you can write a Postmortem under What happened. It shows under the timeline on the incident's page. Saving it emails nobody.
To remove an incident from the page, the history, the feed and the uptime, choose Delete in its menu. It goes to the trash, under Incidents, then Trash, from where you can restore it.