Anomaly detection based on absolute threshold

[martin_schuetz1]

Hi,

I want to open an alert based on a (single) exception for a dedicated service.

Therefore I updated the service settings / error detection and added my custom marker exception.

Then I tried to configure the service settings / anomaly detection --> failure rate to handle occurrences of my exception.

Basically I want to open an alert for 1 minute if there is at least 1 impacted request and automatically close that alert if there is no exception in the following time frame.

What's best practice to configure absolute threshold? If I'm right, both automatic as well as fixed threshold detection are based on relative thresholds (at least I run into problems to automatically close alerts if there's high throughput).

As the exception is not logged log-monitoring is not an option.

Thanks in advance!

[ChadTurner · Accepted solution]

I dont think you can define the duration of an alert, Rather Raise an alert if X amount of breaches, in X amount of time. We recommend using AI, that's what your paying for, and if the case presents itself you can always define a custom event on a case by case basis.