Docs Incidents and alerts
Incidents and alerts
How problems become incidents, how severity is set, and who gets alerted.
One problem, one incident
Every problem a check or an email finds has a fingerprint. For a Flow failure that is the client, the Flow, the element and the error with record IDs and numbers removed. A repeat updates the same incident, adding to its count and last-seen time, instead of opening another.
| What happens | Result |
|---|---|
| A new problem is found | An Open incident |
| It is seen again while Open or Acknowledged | The count, last-seen time and severity are updated. No new alert. |
| It is seen again after it was resolved | The incident reopens and alerts again |
| The check that found it passes, and the problem is gone | Resolved automatically, as “No longer observed” |
| The check that found it fails | Nothing changes. A failed check isn’t evidence that the problem has gone. |
You can acknowledge an incident to show someone is on it, and resolve it yourself.
Collection failures
When a check can’t run, for example because the connection needs re-authorising, the client’s status shows it straight away, in amber. A Collection failure incident, and its alert, waits until the same check fails twice in a row. A blip that fixes itself by the next run doesn’t wake anyone. A changed org ID is raised at once, because it is never a blip.
Emailed Flow failures
A Flow error email opens or updates an incident within seconds.
- With full monitoring, the incident is marked Awaiting confirmation until the next check finds the same failed interview through the API. It is never resolved automatically while awaiting.
- For Flow errors only, nothing can confirm it, so it is labelled Reported by email, not verified through the API. It resolves itself after 7 days with no repeat emails.
Severity
The thresholds are consultancy-wide, in Settings. Defaults are in brackets.
| Problem | Severity |
|---|---|
| Flow failure | High in Production with 5 or more failures in 14 days. Low outside Production. Medium otherwise. |
| Failed logins | Medium with 10 or more in 24 hours, Low otherwise |
| Failed Apex or Bulk job | High in Production, Medium otherwise |
| API use or storage | High at 90% of the allowance, Medium above the attention threshold |
| Collection failure | Medium |
A signal you turn off in Settings raises nothing, and its open incidents are resolved. Collection failures can’t be turned off.
Who gets alerted
- Only new and reopened incidents alert. Updates to an open incident never do.
- Alerts go to the consultants assigned to the client who have Receive alerts ticked.
- Each person chooses, in their preferences, whether to get alerts, the lowest severity they want, and how: email, a Salesforce notification, or both.
- Email sends one message per person per check, listing every new problem. Notifications send one per incident.