On dynatrace documentation I find this information regarding Smart Alerting and to be honest is not clear at least for me the conditions where alerts are evaluated:
"or the generation of alerts, baselines are evaluated within 5-min and 15-min sliding time intervals. The 5-min window serves for quick alerting in case a sufficient number of sample values surpassing a baselines are identified. A 15-min interval is used for generating alerts with higher confidence. However, in case a large amount of sample values is found to be above the baselines within one minute, Dynatrace will generate an alert at this point of time as well"
So sorry but for me is very ambiguous... some times evaluated in 1 one minutes, other in 5 minutes and others in 15 minutes... depending on "sufficiente number of sample values"...
Please. how many sufficient number of sample values in each case?
Solved! Go to Solution.
The documentation you mention above is related to our automatic baseline for services and applications. Here Dynatrace always evaluates a 5min and a 15min sliding window. If your service has a lot of load and the situation is already statistically significant in a 5min sliding window Dynatrace alerts. If the service is low load, the 15min sliding window has more measurements than the 5min window, which means it can alert on significance even where the 5min cannot.
You ask about how many requests are necessary, which is not easy to answer, as it depends on the kind of situation. Imagine an error rate increase of a service with 10 requests per minute. Here a 1% increase would mean that you have 50 requests in the 5min window. 1% would need 100 requests to get one single faulty requests.
The other question you asked about the 1min sliding window is all about infrastructure monitoring and alerting, where we use a 3 out of 5 minute interval for alerting by default.
You can use the feature 'Custom event for alerting' feature to customize this sliding window to your demands.