Skip to main content

Where alerts come from

Most alerts are raised by the platform while it monitors your data. People raise the rest: your team, OpenDataDSL support, the scripts that load your data, and Fusion AI.


Raised by the platform​

RaiserTypeWhenKey
Dataset monitoringcompletenessA delivery is late, missing or incompletedataset/<dataset id>/<date>/completeness
Dataset monitoringquality, criticalA delivery fails its quality or critical checksdataset/<dataset id>/<date>/<type>
Dataset monitoringinitialisationThe dataset's configuration cannot be used, such as a calendar that does not existdataset/<dataset id>/<date>/initialisation
Curve buildsbuildA curve fails to buildcurve/<curve id>/<date>/build
Curve quality checksqualityA built curve fails its quality checkscurve/<curve id>/<date>/quality
Curve exportsexportAn export group does not runcurve-export/<group>/<date>
The platformvarious, category systemA configuration problem, such as an automation target that does not exist

The priority of a completeness alert comes from the dataset's priority (Medium when none is set). A dataset whose alerts setting is off raises no alerts.

note

While an open root cause explains the problem, these raisers record it on the root cause instead of raising an alert of their own.


Alerts that close themselves​

The platform closes its own alerts when the problem goes away:

AlertCloses whenResolution
A late, missing or incomplete deliveryThe data arrives and the delivery is completeDataset is now complete
A delivery on a day the calendar should not have expected dataOpenDataDSL corrects the dataset's calendar from the alert, and the delivery becomes a holidayNo data is expected on the date according to the calendar
A failed curve build or quality checkThe curve builds, or passes its checksThe curve's status
A failed exportThe export runsThe export's status

Alerts raised by people are closed by people.


Raised by people​

WhoHow
Your teamRaise alert on the Alerts screen, to ask OpenDataDSL for help or to track a problem yourselves
Scriptssave ${alert:a} in ODSL, update('alert', ...) in Python, or POST alert/v1/private in the REST API. A script that loads data raises the alert when collection or loading fails
Fusion AIThe raise_alert tool, from a conversation with an assistant. Its alerts are tagged fusion-ai
OpenDataDSLShared alerts about common and public data, and alerts for you about your data

Give an alert a key when the same problem can be raised again (a script that checks something every hour, say): while an alert with that key is open, raising it again counts another occurrence.

An alert raised by a script running in a process is a root cause unless it says otherwise; see Root causes and traces.

note

Alerts are not part of the none plan: on that plan nothing is raised, and people cannot raise or comment on alerts.


Next​

Who sees an alert