I am querying my Druid cluster every two minutes to pull some ad performance data for the current day.
I am basically grouping by two dimensions (ad id and zone id). I get sometime null zone ids for some ad ids (Even if I don’t push null values in Druid for these dimensions via Storm).
Plus this behaviour is not consistent meaning Druid sometime returns a null zone id for one particular ad id but doesn’t return this row when I perform the same query two minutes after.
The other thing I have noticed is that the rows returned by Druid with null zone ids have often metrics near 0 (the number of views associated to the ad and the null zone id seems to be always 0 or 1 for example).
I am wondering if someone already experienced the same issue? I suspect the problem comes from my middle managers, I cannot reproduce this problem when I am only querying my historical nodes.
Thanks for your help!