What to do when a health alert fires

The back office Health panel summarises the platform with one traffic light per item. If something isn't green, here is what it means and what to do.

Backup (Restic)

• Green: the latest backup is under 36 hours old and every target succeeded.
• Amber or red: the backup is late or a target failed. Check the detail in the panel and the restic-backup service log. Until fixed there is no recent backup: avoid risky changes.
• The restore test (once a month) genuinely restores a database and checks it has data. If it fails, the backup is not reliable even though it ran.

Mail and mail alerts

The panel sends an email when something is late. If none arrive, use Mail alerts → Check the connection and Send a test alert; if the test doesn't arrive, the problem is outgoing mail, not the backup.

Server errors and the watchdog

5xx errors are grouped and reported by email. A watchdog alert for a service (for example, a container that restarted) marked “repaired” needs no action, but it's worth knowing how often it happens.

If you don't know what failed

Open Health and click the red item: each explains its cause. Note the time and message before restarting anything.