I have a key service (its a payment service) so I want to know when we are not receiving requests or when the service has low load, so I decided to enable the "detect service load drops" with the follow configuration:
Today we had an incident for about 6 hours because we do not receive any request to the service and the problem for low load was not generated.
Can somebody help me to understand why the problem was not generated for low load?
Solved! Go to Solution.
As soon as you toggle that switch to detect service load drops, the Dynatrace AI starts to learn the typical load of the service over a one-week period. Service outage detection is activated once the one-week learning period has passed. You should be receiving alerts/seeing problems showing up under the Problems tab after the learning period passed. https://www.dynatrace.com/support/help/shortlink/automated-baselining#smart-alerting
This can also happen if the Service has quite low load in general or the traffic pattern fluctuates a lot.
As Moe mentioned, these alerts work based on the 7-day baseline for that service. From observing traffic over the 7-day period, the Dynatrace AI will form a corridor of accepted values based on which it makes predictions and decides whether the observed traffic is expected or unexpected. If the corridor is wide enough, it may happen that 0 requests/min. is low load, but not unexpected low load, therefore not resulting in an alert.
If your Service often hovers at 1-2 requests/min. or normally has periods with 0 requests/min., then a custom alert for request count at 0 requests/min. might be more appropriate than using the "traffic drop" alerts.