Investigate incidents with Maintenance

Maintenance groups retained workflow failures into incidents and supports investigation and reviewed draft fixes. An incident is a starting point for diagnosis, not proof that every grouped failure has one cause.

Collect failures#

Open Workflows → Maintenance and choose the environment. Automatic collection depends on the tenant's enabled settings and configured lookback. Run now lets authorized users select a date/time range of retained failures.

The UI uses the displayed local timezone and submits UTC instants. Check the range carefully when comparing it with logs. A scan can collect a bounded batch; repeat the same range when the UI indicates more retained failures remain.

Turning off collection stops future collection activity; it does not erase existing incidents or automatically cancel explicitly started investigations.

Investigate the evidence#

Open an incident and inspect the affected action, node, environment, status, and occurrences. Classification is a preliminary assessment. Review the recorded evidence and the agent's explanation before accepting a proposed cause.

Investigation requires the appropriate logging and agent permissions. The investigation path is read-only and does not publish changes or replay business requests.

Apply a reviewed proposal#

Where a fix is proposed, review its intended changes before using the approval/application action. Application is constrained to the affected workflow draft. Inspect the resulting draft and run the relevant tests.

A status such as Monitoring means the application job completed. It does not establish that production was released or that the incident is resolved. Use the normal workflow release process, verify the live behavior, then explicitly resolve the incident.

Diagnose one session#

The workflow session's Diagnose session action analyzes saved evidence without changing the workflow. It can be useful even when automatic Maintenance collection is disabled. It cannot recover missing payloads, erased secrets, or details absent from an older snapshot.

Keep the final finding, release reference, and verification outcome with the incident or linked support ticket.