false positive threshold alerts with complex targets about tattle HOT 3 CLOSED

you can fix this in your target with keepLastValue, and you actually want to do it that way, not via Tattle:

diffSeries( sum(keepLastValue(...), sum(keepLastValue(...))

That also filters out spikes if a metric is missing a particular minute.

from tattle.

Thanks.
However adding keepLastValue in every alert target is not natural
I would prefer an implicit fix such as processing data with 1 minute (or n minutes) delay to ensure not working on partial data.
A coworker of mine has done a patch, but we will test it a few days before pushing the patch.

The data I'm working on have never missing samples, however I agree that keepLastValue should be great for that also.

from tattle.

For me it is fixed by #58

from tattle.