Why it matters
A failed run or a batch of rejected rows only gets fixed if someone owns it. When the same problem raises a message on every run, people stop reading them. LakeLogic groups pipeline problems into incidents, shows which ones need a person, and keeps a record of how each one was closed.
What you can do
- See every open incident, and which ones are waiting on a person.
- Take ownership of an incident so the team knows it is handled.
- Read a diagnosis of the cause, and ask Zeus about it.
- Approve a proposed fix, or close the incident with a reason.
How to do it in LakeLogic
Incidents come from the telemetry your pipelines send, such as failed runs and rows held in quarantine by a quality rule.
-
Go to Operate → Incidents. The number beside it counts every active incident.
-
Pick a view:
View Shows Open Every active incident. Needs action Incidents waiting on a person. Assigned to me Incidents you own. Awaiting approval Incidents with a proposed fix to approve. Closed Resolved incidents and why they closed. -
Open an incident. It shows when it was detected, its domain and system, and its impact: rows blocked, trusted rows and quarantine rate.
-
Acknowledge it to record that a person owns it.
-
Read the diagnosis. Its heading says where it came from: Zeus diagnosis when Zeus answered, or Rule-based diagnosis when no AI answer was available.
-
Use Ask Zeus to discuss the incident with Zeus, scoped to it. See Ask Zeus. The actions menu can also refine the diagnosis, create a Jira ticket when Jira is connected, or draft an email to a data supplier.
-
Resolve it:
- When a fix is proposed, choose Approve fix. When a fix pull request has been drafted, the button opens that pull request as well, and it goes through the same checks as any contract change.
- Otherwise choose Close incident and pick a reason: Fixed at the source, Re-run passed, Contract changed, Expected data accepted, Duplicate of another incident, Not an issue, or Other (details required).
What you get
- A closed incident shows its closure reason in place of its status.
- The Closed view keeps every resolved incident.
- Knowledge, under Incidents, keeps diagnoses saved for reuse.
- Alerts about an incident link straight to it, so email, Slack, Teams and the in-app bell open the same page.
Good to know
- Impact figures that were not reported are shown as not recorded, not as zero.
- A run can succeed and still open an incident, for example when a rule rejects rows.
- Late or missing data is tracked as a service level, not as an incident. See Know when data is late or missing.