Why it matters
A pipeline can succeed and still deliver data that is stale, short or unusable. Consumers notice first, and trust drops. Alerts that fire on every check get muted, and then the real problem is missed. LakeLogic tracks each objective, treats a repeated failure as one breach, and sends each alert to the channel that should see it.
What you can do
- See which datasets meet their freshness, volume and quality objectives.
- Tell objectives that were never checked apart from those that passed.
- Get one alert per system run instead of one per table.
- Send team events to Slack or Teams, and choose your own personal alerts.
How to do it in LakeLogic
Check service levels
-
Go to Operate → Service Levels → Overview to see results, breaches and owners.
-
Each objective (Freshness, Volume, Quality target) is in one of these states:
State Meaning Not configured No objective is declared. Not evaluated Declared, but no result has arrived. Stale The last result is too old to describe now. Breached The last check failed. Met The last check passed. A freshness pass only stands while it is recent. With no window declared, a result older than a day no longer counts.
-
Open Evaluation history to see each check behind the summary.
Route alerts to the team
- Connect Slack or Teams. See Connect your tools.
- An organisation admin opens Admin → Notifications (Notification Routing) and chooses which events reach the shared channels. These settings apply to the whole team.
Choose your own alerts
- Go to Settings → My Notifications.
- Choose which events reach you in-app, by email or by personal webhook, and how often.
Review what was sent
Operate → Notifications shows your alerts and why you received them. Delivery activity shows every delivery attempt across all channels.
What you get
| Alert | When it is sent |
|---|---|
| Pipeline run summary | One per system run, once every enabled contract has reported or the run goes quiet. A run where everything passed sends nothing. |
| Pipeline failure | At once, for the first failed table of a run in a production environment. |
| Service level breach | Once per breach, not once per check. A breach stays open until a passing check closes it. |
| Volume anomaly, missing partition | When a load is unexpectedly small, or an expected partition never arrived. |
Good to know
- Breach alerts are off by default for each organisation, so an estate that has not reviewed its objectives is not flooded.
- A breach that is no longer checked stays open and is marked stale. It is never shown as recovered.
- Breaches do not open incidents. Failed runs and quarantined rows do. See Find and fix a data incident fast.
- One person turning off Slack in their own settings does not silence the shared channel.