Docs » Scenarios for finding and resolving infrastructure problems using alerts and detectors » Scenario: Kai sets a Min Delay for detectors to account for sparse data

Scenario: Kai sets a Min Delay for detectors to account for sparse data 🔗

In Splunk Observability Cloud, the site reliability engineering (SRE) team at Buttercup Games has set up a detector called Service error that monitors Buttercup Games services and alerts when the error rate of any service exceeds 5%.

Despite the alert, Buttercup Game has been receiving complaints from customers about errors coming from the productcatalog service.

Kai, an SRE on the team, investigates productcatalog and finds out that the service has a low request rate and rarely sends a data point for errors, sometimes only once a day. Because of this, the analytics engine can’t keep track of the increased delay of the sparse error-tracking MTS. Its data points are too delayed to be included in the Service error detector, so the detector can’t generate alerts on the productcatalog service.

In order to account for delayed data points from productcatalog, Kai configures a Min Delay value for Service error by making an API call to the /detector endpoint. The Min Delay configuration forces the detector to wait a set amount of time before running the computation.

Kai charts the lag from the sparse MTS and determines that it is usually less than 30 seconds but has never been higher than 1 minute, so they set a Min Delay of 1 minute for Service error.

Summary 🔗

By setting a Min Delay threshold for Service error, Kai has successfully included sparse data and can now catch errors from the productcatalog service before they affect customers’ experience.

Learn more 🔗

For more information on Min Delay for detectors, see Min Delay.

For more information on the Detectors API, see the Detectors API reference .