I’ve discovered some unexpected behavior with AppDynamics in regard to how health rules are calculated when you have a fluctuating threshold.
Metric values for each minute are compared against the *current* threshold - not what the threshold was at the time the metric value was captured. This can become evident when you use baselines and standard deviations as your thresholds and you look back some number of minutes. Say you have a health rule that looks back 15 minutes. If you evaluate the health rule at 10:01, you're using the baseline from the 10:00 hour to evaluate metric values from the 09:00 hour.
Here's a screenshot to further explain:
Here’s the alert I mention in the above screenshot:
(I've edited out the server name from both screenshots)
Is this how health rules using baselines and lookbacks work across the board? Or is it only like this for Database health rules? And is this intended behavior? Is there any way to work around it?
... View more